May 13, 2019

THE FUTURE COMPUTED

AI and Manufacturing

by

Greg Shaw

Foreword by Çağlayan Arkan

1
 Published by Microsoft Corporation
Redmond, Washington U.S.A.
2019
2 3
 TABLE OF CONTENTS

01 FOREWORD

11 EXECUTIVE SUMMARY

CHAPTER 1: THE FUTURE OF AI

19 AND MANUFACTURING

Boundless Collaboration: Digital Transformation Means

Cultural Transformation
Seizing the Moment: Capabilities as a Competitive
Advantage
Leveraging the Data Estate: Turning Insights into Action

CHAPTER 2: AI AND THE

47 MANUFACTURING WORKFORCE

A New Relationship with Technology: Creating a Safer and

More Satisfying Workplace
The Talent Supply Chain: A Shared Agenda for
Skills and Employability
Working Together

CHAPTER 3: FOSTERING RESPONSIBLE

69 INNOVATION

Ethical AI in Manufacturing
New Rules for New Technology
Wheres to from Here

109 CHAPTER 4: THE WAY FORWARD

First published 2019 by Microsoft Corporation AI Maturity and Progressing Your Journey
One Microsoft Way
A Path to Innovation
Redmond, Washington 98052
117 CONCLUSION
© 2019 Microsoft. All rights reserved
4 iii
3
FOREWORD Indeed, what has emerged is a whole new branch of robotics,
developing what are called “collaborative robots” that are working
safely alongside people. FORWARD

These “cobots” are replacing many of the traditional robots kept in

cages or behind safety screens and can be programmed to perform
It has been a long and eventful journey for me from my time as a point
a wide variety of tasks, making them a far more attractive option for
guard in the Turkish professional basketball league in the 1980s to my
manufacturing companies of all sizes and for their workers in all manner
role leading Microsoft Corp.’s Manufacturing & Resources Industry
of scenarios.
Group today. And while I have traded my sneakers for a computer, the
experience of what teamwork can achieve has never left me. Today I am Yet, despite their impact on making manufacturing more efficient
part of a different team—one committed to helping our manufacturing and safer, and reducing the number of tedious and repetitive roles,
customers adopt, embrace, and lead the digital transformation of their automation and robots have been getting their fair share of criticism.
organizations in an era of AI. Some critics speak about the need to have “robot-proof jobs”1 or to tax
them like employees if they take the job from a human. 2 But from our
The times could not be more exciting for Microsoft and the
experience, and the experience of many of our customers, the reality
manufacturing industry, especially when you think of the opportunities
appears more beneficial than bleak.
that AI will present in defining both what manufacturers create and how
they create it. Returning to the theme of sports, technology is enhancing human
and team performance, and even creating more jobs. For example,
This excitement is not surprising given that manufacturing is often
professional teams around the world have analytics departments that
a relatively early adopter of new technologies. As one of the first
employ data scientists to analyze and support players’ performance.
industries transformed by the Industrial Revolution in the 18th century,
Similarly, new roles are being created in manufacturing to develop new
manufacturing has already established itself as a pioneer in the Fourth
technologies and leverage AI. And this comes amid an existing job
Industrial Revolution in this century.
shortage in manufacturing.
Today manufacturing has a strong association with AI, especially the use
The Manufacturing Institute and Deloitte estimates that the United
of automation. But the automation that we have typically associated with
States alone may have as many as 2.4 million manufacturing jobs to fill
manufacturing in years past was often production lines, heavy machinery
between now and 2028, with around 500,000 unfilled jobs in the sector
and, more recently, industrial robots. These robots were expensive
today. 3 And there continues to be a shortage within the manufacturing
and bulky machines that were mostly used in large factories and out
workforce, with skilled trades and manufacturing jobs ranked among
of reach for many small and medium-sized manufacturers because of
the hardest roles to fill for the past ten years according to the 2018
limited budgets or factory floor space. In recent years, however, thanks
ManpowerGroup global Talent Shortage Survey. 4 What is more, as the
to advances in AI technology, these robots have become smaller and
second wave of the Baby Boomer generation moves into retirement, this
far cheaper and are leading a very different interaction and relationship
trend will only worsen. 
between humans and technology on the factory floor.

1 2
It’s hard to be certain about the impact of AI on manufacturing. And Microsoft has long been a technology partner in manufacturing, and,
while automation might displace some jobs and some roles, there are over time, we have learned how important it is to bring customers into
more jobs and opportunities being created as automation and AI helps our roadmap planning. FORWARD

drive a manufacturing resurgence in many regions. Indeed, PwC forecasts

Today our engineers share their roadmaps with manufacturers like
that AI will contribute to a 14 percent increase in global gross domestic
Komatsu, Rockwell, John Deere, Airbus, and many others to help shape
product by 2030, or around US$15.7 trillion, as the technology increases
and reshape priorities.
productivity, product quality, and consumption. 5
Beyond this roadmap sharing, we are also co-creating with customers.
Other research suggests that AI could double annual economic growth
For example, Microsoft’s Power BI was shaped, in part, through work
rates in mature economies by changing the nature of work and creating
with airplane engine manufacturers like Rolls Royce, and our work with
a new relationship between humans and machines. 6 In Germany alone,
industrial conglomerate thyssenkrupp has helped design our Internet of
one government-sponsored study predicts that between 2018–2023, AI
Things (IoT) priorities.
will add €32 billion to the country’s manufacturing output: This equates
to about one-third of the entire growth expected in this sector over the But we are not just a technology partner with our manufacturing
same period.7 customers—we are a manufacturer in our own right. For over 30 years
we have been producing peripherals like keyboards and mice, gaming
The future of manufacturing and manufacturers’ ability to improve key
consoles, personal computers, and mixed reality headsets along with
areas, such as worker safety and sustainability in directly dependent on
other related products.
AI and technology. We hear that firsthand from customers and see the
value they are already delivering. We understand the complexities of a global supply chain and the
opportunities of applying our own technology solutions to our own
Similar to a coach, my job today is not simply about helping customers
manufacturing operations to help us be more efficient, sustainable,
with their transformation to an AI company; it is also about releasing
and competitive.
this potential of the teams that make up our customers and the broader
economy. That is what inspires me and our entire team every day. The insights we have drawn from our own and our customers’
experiences reveal that we have only scratched the surface of the true
Microsoft’s approach to help this transformation is to start with the
impact of AI to the manufacturing sector, its workforce, and society
business outcome our customers want. Only then do we identify the
more broadly.
necessary insights from their own data estate and what models will help
realize the transformation impact they seek. To expand on this understanding, we went out and interviewed many
of these manufacturing leaders, including our own operations team, to
We want customers to be agile and nimble by breaking down data silos
look beyond the hype of AI. We wanted to discern what is real, what is
within their organizations and creating a cognitive supply chain where
possible, what is worrying, and what stands in the way.
AI can make 30 to 40 percent of decisions, releasing time to do the things
that humans do best.

3 4
The Future Computed: AI and Manufacturing picks up where the
inaugural volume of Microsoft’s The Future Computed: Artificial
Intelligence and its Role in Society left off. This book is a deeper dive into
the challenges and opportunities our small, medium, and large customers
face across a broad spectrum of manufacturing businesses. It introduces
you to the people we are proud to work alongside of and shares their
inspiring stories of the progress of AI.

This is the story of inspirational human achievement, winning teams, and

organizations. This is the story of companies who are often dreaming big,
sometimes starting small, but always aiming to move fast.

ÇAĞLAYAN ARKAN
Global Lead
Manufacturing & Resources Industry Group
Microsoft Corp.

ÇAĞLAYAN ARK AN

5 6
WHAT AI MEANS TO MICROSOFT aircraft engines. It’s also the technology behind Danish brewer
Carlsberg experimenting with whether AI can help predict how new
beer varieties will taste. 

But AI systems are only as good as the data from which they learn.
We work with our customers to create AI solutions that will help
When you ask someone what they understand by the term “AI,”
harness their data and create better insights for their organizations.
the response is often framed by how the technology is portrayed in
science fiction novels or Hollywood blockbusters. But, as with many When we add AI capabilities to existing products, or release new
technologies, the reality of AI is less sensational and more practical products infused with AI, they are often rooted in discoveries from
than our imagination is led to believe. Microsoft’s research labs. 

Put simply, AI is a machine’s ability to recognize patterns, sounds, Microsoft researchers have led breakthroughs in fields including
images, and words, and to learn and reason over data—to infer—in image recognition, machine translation, speech recognition, and
ways that are similar to what people do. It’s a set of technologies machine reading comprehension, which uses AI to read, answer,
that enable computers to understand and interact with the world and even ask questions. Those breakthroughs then make their
more naturally and responsively than in the past, when computers way into products such as Azure Cognitive Services, a set of tools
could only follow preprogrammed routines. developers can use to add AI capabilities such as image recognition,
translation, and visual search into their products, and the bot-
AI isn’t new, and it certainly isn’t new to Microsoft. In fact,
building tools in Azure Bot Service. 
computer scientists at Microsoft and elsewhere have been working
on machine learning and AI technologies for decades, and we Taken together, these advantages are enabling Microsoft and our
are now witnessing the realization of this work in a wave of customers to create products and services that use AI to better
breakthroughs that promise to open a new era for AI applications in understand, anticipate, and respond to people’s needs.  
business and society.
One example of AI in use is Cortana, Microsoft’s intelligent digital
This move from the computer lab into mainstream products is assistant. Cortana can do everything from scheduling a meeting
thanks to a set of contributing factors: the massive computing for you, to reminding you to follow up on a specific commitment
power of the cloud; the availability of enormous datasets that can be you made in an email. Cortana is able to do things like give you
used to teach AI systems; and the breakthroughs in developing AI reminders because it’s been trained using massive, anonymized
algorithms and improving AI methods such as deep learning.   datasets to understand why a specific email is important. Cortana
will then learn specifically what emails might be important to you
At Microsoft, we have advantages in every one of these areas. 
based on your behavior and preferences.
The immense computing power of the Azure cloud offers customers
far more than just a place to store data. It allows companies like
Rolls Royce to discover actionable insights around fuel usage,
predictive maintenance, and stopping unscheduled delays for their

7 8
But beyond digital assistants, Microsoft’s vision is to democratize AI through an earpiece. Seeing AI allows people with low vision or
even further. To realize this, we envision a “computing fabric” for blindness to navigate the world with greater independence.
everyone, one that blends the intelligent cloud with the intelligent
Finally, with the world’s population expected to grow by nearly 2.5
edge—where data is analyzed and aggregated closest to where it
billion people over the next quarter century, AI-driven farming
is captured in a network. We envisage a world where this cloud
offers hope that we’ll be able to grow enough food to meet the
and edge computing is coupled with a multi-sense, multi-device
increased demand. Our FarmBeats project is using the power of the
experience that seamlessly integrate speech, gestures, and the gaze
cloud and machine learning to improve agricultural yields, lower
of our eyes. This is a world where mixed reality—virtual worlds with
costs, and reduce the environmental impact of farming.
digital twins—will become the norm.
At Microsoft, we believe that these stories provide just a hint of
“We are creating new experiences and enabling digital
the unprecedented opportunities that AI offers to address pressing
transformations for every customer,” Microsoft CEO Satya Nadella
challenges, drive progress, and transform how businesses operate
told shareholders at the 2018 annual meeting. “We’re leading
and people work.
the field of AI research, achieving human parity with object
recognition, speech recognition, machine reading, and language
translation. But, most importantly, we are focused on democratizing
these AI breakthroughs to help organizations of all sizes gain their
own competitive advantage, because of AI.”

But AI is not simply about productivity and efficiencies; it is about

providing resources and expertise to empower those working to
solve humanitarian issues and create a more healthy, sustainable,
and accessible world.

One example is from our team at Microsoft Research in the U.K.

who are working with a group of oncologists on a project called
InnerEye that uses AI to mark the boundaries between healthy
tissue and tumors in CT scans. Today this is a time-consuming job
that is done by hand—pixel by pixel across in hundreds of layers of
scans. Thanks to InnerEye, the job can be accomplished in seconds.

Another powerful example is a tool called Seeing AI, which was

developed by Saqib Shaikh, a Microsoft engineer who lost his sight
at age seven. This powerful mobile app reads signs and documents,
identifies currency and products, recognizes friends, and interprets
peoples’ facial expressions so it can describe what it sees to the user

9 10
 EXECUTIVE SUMMARY

EXECUTIVE
SUMMARY

When Bill Gates and Paul Allen founded Microsoft four decades ago,
their goal was to democratize the benefits of software for computing—
then largely restricted to enormous, expensive mainframes—to everyone.
Their vision was a computer on every desk and in every home. Today,
Microsoft is aiming to do something similar with AI—to make AI
development and its benefits available to everyone.

Our AI systems are designed to empower people and to augment their

capabilities, and we’re committed to making sure our AI tools and
technologies earn their trust.

To help advance the conversation on how AI can empower people while

fostering trust in the technology, in January 2018 Microsoft published
The Future Computed: Artificial Intelligence and its Role in Society. The
book represents our perspective on where AI technology is going and
the implications for society in a future where the partnership between
computers and humans will be more pronounced than ever before.

Microsoft President, Brad Smith, and Executive Vice President of

Microsoft AI and Research Group, Harry Shum, wrote in the book’s
foreword that AI will enable breakthrough advances in every industry
while raising complex questions and concerns. They sought to answer
some fundamental questions, such as: How do we ensure that AI is
designed and used responsibly? How do we establish ethical principles to
protect people? How should we govern its use? And how will AI impact
employment and jobs? Ultimately, the question they posed is not what
computers can do, but what computers should do?

In this companion book—the first of a series to explore AI, the future

of the workforce, ethics, and policies related to individual industries—

11 12
we delve into manufacturing, an enormous global ecosystem and the Listening to our customers and these stakeholders, we discovered
backbone of our human economies. It’s a sector that throughout history six themes:
has embraced new technologies while shaping the nature of work and EXECUTIVE
SUMMARY
1. Manufacturers are already seizing the AI opportunity. AI
creating enormous economic and societal impact.
is not simply about productivity. It’s also about helping reinvent
When we first conceived this book, we had three audiences in organizations. AI is about workplace safety and health, predictive
mind: manufacturers who are curious about the AI journey of their maintenance, process efficiencies, intelligent supply chains, up-
colleagues; policymakers and regulators who are examining AI’s societal time, higher value, and higher-quality products. These are the
implications; and our commercial partners who want to stay abreast of drivers of AI design and implementation for the manufacturers
both the technological breakthroughs and the policy possibilities. we spoke with. In addition, as a global manufacturer, Microsoft is
using AI to drive its own transformation, and we are sharing those
In scores of interviews with manufacturers, workforce experts, unions,
insights with our customers.
and policymakers, we heard time and time again that manufacturers and
workers alike are embracing this AI journey. In the pages that follow, we 2. Central to digital transformation is cultural transformation.
hear customers of Microsoft describe that journey, including their views Strong leadership combined with engaging workers at all levels
that this technology has profound implications for their business models, in the process is essential. In order to optimize AI’s value, the
their workforce, and their responsibilities as ethical organizations. entire organization must work together to embrace change, break
down silos, and create a seamless information supply chain inside
We hear from labor organizations that they are encouraged about how
companies and leverage their full data estate.
AI can improve worker safety and job satisfaction, but they are pragmatic
about the pace of technology change. They recognize the need to be 3. Those closest to the workforce, the managers and leaders
planning for the impact on workers, especially by building adaptable inside manufacturing operations, are often the most sensitive
skills to meet the opportunities and be more resilient to change. to AI’s impact on their workforce. Their focus is to create a
We hear from policy makers that they are excited by AI’s potential to better company and more opportunity, including a safer work
help drive jobs and economic growth, but they are anxious to have the environment, fewer repetitive and unsatisfying jobs, and increasing
right policy environment to help realize the technology’s potential in a productivity. They put their people first and so are eager to adopt
responsible way. technologies that have positive impact on workers.

4. There will be disruption and dislocation, and we need a new

pipeline of talent. While there is positive sentiment about the
opportunities for the workforce, manufacturers are also concerned
about talent shortages, short-term disruption, and attracting the
next generation of bright students. Jobs in manufacturing will
require new skills and new capabilities, and this necessitates a new
partnership for skills and workforce development with technology
providers, industry, government, learning institutions, and
labor organizations.
13 14
 5. Next-generation policies and laws are needed for next- So, just as Microsoft is seeking to better understand the industry policy
generation technologies. Our customers are already versed in landscape, we are also encouraging our customers to be more engaged
many of the ethical issues that are associated with AI—security, in the development of digital policies. We want to bring together leaders EXECUTIVE
SUMMARY

safety, reliability—but are looking for more guidance on how to use across private and public sectors, civil society, employees, and labor
AI ethically and responsibly. As manufacturers, they are using AI in organizations to co-create the right policy and ethical frameworks where
their production processes and infusing AI into their products, and industry opportunity and human potential can both flourish.
they are therefore looking for guidelines that will help anticipate
Second, all of us need to prepare for the impact on the workforce and
potential issues and ensure responsible innovation. Similarly,
build a supply chain of talent to help new workers coming into the
regulators are eager to remove technology hurdles and encourage
workforce acquire the new skills they will need, develop workers who will
the adoption of AI technologies that will promote worker safety,
have to transition to new jobs with the same employer, and support those
create more jobs, and help national competitiveness.
whose roles will be eliminated and will need new jobs elsewhere in
6. AI is a journey, different for everyone. While many of our the economy.
leading customers have embraced AI, there are many who are just
But, finally, what this really means is we must keep engaging. Just as
beginning their journey. Microsoft has developed a guide to help
we outlined in the original Future Computed, we’re all going to need
them on their AI journey and show them how to best leverage
to spend more time talking with, listening to, and learning from each
their data estate. Implementation ranges from improving reporting
other. No single company has all the answers, and this book doesn’t
through data analytics and business intelligence to “deep” AI
seek to present all the issues that an industry as broad and as diverse as
capabilities through machine learning applications.
manufacturing is facing. This book is just part of our ongoing engagement
But what do we take from these themes and how do we move forward and dialogue to surface issues and build coalitions of interest to address
with these insights? them. We continue to look forward to working with people in all walks of
life and every sector for building a foundation for AI that transforms not
First, it’s becoming increasingly clear that society needs a long-term and
just businesses, but transforms lives.
multi-stakeholder approach to realize industry transformation potential
through AI. As the lines between industry policy and technology policy
continue to blur, coupled with the progress of manufacturing companies
becoming digital companies, the need for our customers to engage in the
development of the digital policy agenda is more important than ever.

15 16
THE AI POTENTIAL: THE DATA

THIS INFOGRAPHIC IS BASED ON MICROSOFT ANALYSIS OF

THIRD-PARTY DATA. SOURCES INCLUDE:
(PWC)

1. Manufacturing the Future: Artificial intelligence will fuel the next wave of growth for industrial equipment company.
Accenture 2019 https://www.accenture.com/t20180327T080053Z__w__/us-en/_acnmedia/PDF-74/Accenture-Pov- 8. Manufacturing the Future: Artificial intelligence will fuel the next wave of growth for industrial equipment company.
Manufacturing-Digital-Final.pdf#zoom=50 Accenture 2019 https://www.accenture.com/t20180327T080053Z__w__/us-en/_acnmedia/PDF-74/Accenture-Pov-
Manufacturing-Digital-Final.pdf#zoom=50
2. McKinsey – Digital Manufacturing Capturing Sustainable impact at scale (June 2017)
9. Food and Agriculture Organization, How to Feed the World in 2050 (2015), www.fao.org/fileadmin/templates/wsfs/docs/
3. IDC FutureScape: Worldwide Operations Technology 2017 Predictions Jan 2017 Doc # US42261017 Web Conference By: expert_paper/How_to_Feed_the_World_in_2050.pdf
Lorenzo Veronesi, Marc Van Herreweghe.
10. OECD, Water Outlook to 2050: The OECD calls for early and strategic action, (May 2012) www.globalwaterforum.
4. https://futureoflife.org/national-international-ai-strategies/?cn-reloaded=1 org/2012/05/21/water-outlook-to-2050-the-oecd-calls-for-early-and-strategic-action/

5. The Economist Intelligence Unit, Intelligent Economies: AI’s transformation of industries and society, (July 2018) https:// 11. World Economic Forum, Future of Jobs report 2018 (September 2018) http://reports.weforum.org/future-of-jobs-2018/
eiuperspectives.economist.com/technology-innovation/intelligent-economies-ais-transformation-industries-and-
society?utm_source=Organic%20Social&utm_medium=Twitter&utm_campaign=Microsoft%20-%20Intelligent%20 12. IDG Communications, Top 5 cybersecurity facts, figures and statistics for 2018 (January 2018) https://www.csoonline.com/
Economies&utm_content=Briefing%20Paper article/3153707/security/top-5-cybersecurity-facts-figures-and-statistics.html

6. World Economic Forum, Future of Jobs report 2018 (September 2018) http://reports.weforum.org/future-of-jobs-2018/ 13. IDG Communications, Top 5 cybersecurity facts, figures and statistics for 2018 (January 2018) https://www.csoonline.com/
article/3153707/security/top-5-cybersecurity-facts-figures-and-statistics.html
7. American Council for an Energy Efficient Economy, How Smart Buildings Save Energy, (Nov 2015) https://www.buildings.
com/article-details/articleid/19537/title/how-smart-buildings-save-energy 14. Cybersecurity Tech Accord, https://cybertechaccord.org/about/

17 18
CHAPTER 1

THE FUTURE OF AI
AND MANUFACTURING

19 20
THE FUTURE OF AI AND MANUFACTURING BOUNDLESS COLLABORATION:
DIGITAL TRANSFORMATION MEANS
CULTURAL TRANSFORMATION THE FUTURE
OF AI AND
MANUFACTURING

The Ups and Downs of AI

Whether standing on a plant floor or in a cavernous warehouse, The sound of a phone ringing in the middle of a weekend night awakened
manufacturers light up when talking about efficiency, productivity, Patrick Bass from his deep sleep. As a young elevator technician, Patrick
and performance. They lower their eyes and sigh when they speak of was not completely surprised because it had fallen to him to be on call
downtime and legacy machines. But the potential of smart factories and if a customer needed help. What did surprise him was the voice on the
their own individual roles in leading a small part of a much larger AI other end.
revolution is evident in every conversation with manufacturers across It was an elderly woman who was trapped in her home elevator. She
the world. was understandably upset and gave him instructions for how to get into
Very often those we interviewed are engineers—or entrepreneurs with her house. He threw on some clothes and sped over immediately. The
engineering inclinations—who rose through the ranks. They didn’t set situation he discovered was troubling. The woman’s robe had been caught
out in life to lead an AI revolution, but their deep knowledge of the inner in the elevator system, and she was only partially clothed. The reason
workings of their manufacturing processes thrust them onto the stage of she was upset had only a little to do with her own predicament; she was
next-generation technologies. They are a bridge between the old and mostly upset that she had to disturb Patrick in the middle of the night.
the new. “Her sincerity is unforgettable,” he said. “It helped me work through the
The following chapter is about sharing their stories and what they can uncomfortable situation.”
teach us about what Microsoft CEO Satya Nadella calls “tech intensity.” Patrick, who today is the CEO of thyssenkrupp North America, is
This is where the potential for companies and countries to jump-start indicative of so many manufacturers we spoke with—brilliant engineers
their growth by not just adopting technology, but by building their pioneering new technologies and sensitive leaders with profound
own technology and skills too. One way to think about tech intensity emotional intelligence.
is creating the right environment for the fusion of cultural mindsets
and business processes—a fusion that rewards the development and Born on a dairy farm near Burlington, Wisconsin, his mother had been
propagation of digital capabilities to create boundless collaboration confined to a wheelchair since age 13. “She wasn’t supposed to have kids,
and results in new insights and predictions, automated workflows, and so I am a mom’s boy.”
intelligent services. He was able to care so perceptively for the woman caught in the elevator
Understanding tech intensity is best illustrated through the stories of our because he had been caring for his mother for as long as he could
customers—stories that span meaningful results such as collaboration, remember. Helping people was a natural part of his life.
competitive advantage, and creating better data supply chains. Patrick loved systems and understanding how things work. He knew
milking cows was not his long-term future, and his mother and

21 22
grandmother were intent on him going to college. He got an engineering
degree and always dreamed of designing cars. THE FUTURE
OF AI AND
MANUFACTURING
His first job was working with an original equipment manufacturer to
design disability products for automobiles. In time, Patrick moved on to a
company called Mobility Unlimited. As fate would have it, his family also
had an electronic business which had enabled him to get an electrician’s
certificate. The elevator mechanics at his new employer needed a certified
electrician to sign the license, and so he became an elevator helper,
simply handing tools to the mechanic. He fell in love with the job and
began to service the products Mobility Unlimited didn’t want to service.
He once worked on a service elevator dating back to 1917 that employed
sheets of leather with water hydraulics to pull cables from a restaurant
below ground to the sidewalk above. There was no manual to read so he
had to figure things out on his own. “I’ve been here before and you are focused on [the] wrong things,”
he told them.
He went from helper to full lead mechanic in a year. Today Patrick
has spent 26 years in the elevator industry, working on all facets from An executive called for a break to conference with Patrick, and when the
designing and manufacturing to servicing and selling elevators. meeting reassembled, Patrick was put in charge of the project. But that
wasn’t all.
He started at thyssenkrupp as Product Engineer “Number 1” and
progressed up the chain of command from there. He had gone from a “They can either reshape it or kill it,” the executive declared.
small elevator company to a very big one. Eventually he was assigned to
Fate again struck. Just as Patrick was reimagining the project, he went to
the research group and became fascinated with the data generated by
see the new CEO of Microsoft, Satya Nadella. Satya had been in the new
elevators. He pushed himself and the team to understand how they could
role only a few months. They met and realized they shared a vision about
extract data from elevators to make predictions about maintenance
what data and AI could be.
and safety.
Patrick said, “Satya looked at us and said, ‘Let’s do this.’ We will do a
“We failed miserably,” he said. “It left a tarnish, but we regrouped and
proof of concept together, but we’re going to do this in three months, and
worked through it.”
we need you to keep up.”
Years later, he sat in a corporate meeting and learned about a “black box”
Like the proverbial phoenix that rises from its own ashes, thyssenkrupp’s
project in the field that involved streaming elevator data to a mechanic.
Max—Maximum Uptime, All the Time—was born. It was the elevator
The presentation focused on the device, not the data, the cloud, or AI.
industry’s first real-time, cloud-based, predictive maintenance solution.
He spoke up.
With more than 12 million elevators moving one billion people every
day, uptime is important to any elevator, whether it’s a thyssenkrupp

23 24
elevator or not. But digital transformation had come to thyssenkrupp Silos with a View
quickly—and digital transformation means data is available to everyone
Thyssenkrupp’s story of cultural transformation is not isolated. The THE FUTURE

all the time, in real time, anytime, and on any device for the purpose of OF AI AND
MANUFACTURING
collaborative work Patrick and his team led to bring new products and
empowering people and business.
services to life is mirrored at another of our customers, Jabil.
“One source, one truth,” Patrick recalled. “Everything is a system. People
Jabil is a global manufacturing solutions provider with more than 200,000
systems are the most dynamic and most rewarding. My job as a leader is
people across 100 facilities in 29 countries. One of those facilities is
to help empower people.”
in Tampa Bay on Florida’s west coast, and this is where we meet Matt
Since the initial success, thyssenkrupp has taken this cultural Behringer, Jabil’s VP of Operational Technology.
transformation to help other customers. For instance, one auto
Matt bristles when you ask if AI can tear down silos to make the business
manufacturing customer of theirs was struggling with one specific
more connected and agile. He will correct you. In fact, he argues, every
operation on the assembly line—sealing a door. It was a costly, time-
business needs some silos to ensure functional depth. Those silos,
laden job, but it was also a job that AI could help improve. thyssenkrupp
however, need windows to provide adjacent businesses and services with
realized the customer was not scaling data beyond a single factory in
the visibility they need to collaborate and optimize for customer needs.
order to get bigger impact across the company. The factory simply said
“no” when asked for its data. “Sure, you are corporate but you’re not Matt is leading Jabil’s “Factory of the Future” initiative. The vision is to
taking my secret sauce,” the factory manager said. create an autonomous, adaptable factory.

Over time, though, the manufacturer was able to get everyone on board “We need things connected and interacting so that when change occurs,
to share data. Stakeholders came to understand the data was not going to the technology understands what’s going on and can adapt to that
be used against them. Instead, they shared data to build a digital twin—a change,” Matt said. “Equipment, people, and technology—all of it.”
virtual assembly line—to come up with a better solution. They now have
To accomplish this, they’ve moved from organizing by traditional
a brand-new processing system that takes not just one piece of data, but
internal functions to cross-functional workstreams. Automation across
all the data necessary to design a new system. Patrick and team were able
the entire organization is the goal. A connected system will correct itself
to go to the customer and present a new solution. Overnight, they had
before there is a problem. Is the system drifting or are there human
hundreds of millions of dollars in new revenue from top-tier
issues? In the past, sensors and cameras were not far enough along, but
automotive producers.
today they are.

“Everything is a system. “Don’t make a defect, fix it now,” Matt said. “Do it automatically rather
than using charts to diagnose later.”
People systems are the most
Even in an era of automation and AI, Jabil sees humans continuing to
dynamic and most rewarding.” be central to the equation. And while people may be repurposed and
retrained, giving humans real-time feedback to make improvements
is essential.

25 26
 Matt should know. Born in Illinois on a dairy farm, he moved as a
teenager to Florida and after college became a general contractor on THE FUTURE

boats. This is where he met one of the founders of Jabil who noticed his OF AI AND
MANUFACTURING

technical and entrepreneurial skills and quickly hired him. Starting in

production in a newly constructed building, Matt moved up quickly in
the manufacturing operation. At the time, Matt said he couldn’t even
spell IT; but Jabil asked him to learn information technologies and over
time his math and engineering skills absorbed that as well. He became
fascinated with how the company could empower people to do their jobs
better. Today he is the vice president in charge of operational technology,
including Jabil’s Factory of the Future initiative.

“We are not taking away silos, but we are putting windows in them so
we can compound the value proposition—aim everyone in the same
direction. We need silos but we need them harmonized.”

The experiences of thyssenkrupp and Jabil tell us that cultural changes

are as important as the AI technologies in helping empower employees
and make data-driven decisions. These organizations are focused on
adopting a data culture and continuing to utilize prioritized strategic AI
initiatives to disrupt their industry and create new business models
and streamline operational processes. Next, we introduce you to some
equally inspiring organizations that are harnessing AI to carve out
competitive advantage.

SEIZING THE MOMENT: CAPABILITIES AS A

COMPETITIVE ADVANTAGE

Paper Planes

Manuel Torres was born and raised in the Spanish city of Murcia. An
entrepreneur and self-taught engineer, Torres has made the long (and
unlikely) journey in manufacturing from pulp and paper, to aerospace, to
AI. Early in his career, during the 1973 oil crisis, prices for raw materials
like paper increased nearly 100 percent and scrap became an important
engineering challenge.

27 28
The paper company Manuel worked for, like the industry itself, was and tails. The move into aerospace was propitious. Not only would the
struggling with outdated, inefficient processes. There was a lot of scrap, industry need MTorres’ capabilities in the near term, they would also THE FUTURE

which meant a lot of waste and lost profit. be needed for cutting-edge new endeavors. OF AI AND
MANUFACTURING

Torres had an idea for automating the splicing of one massive roll of paper The Seattle Times reported in early 2019 that Boeing is investing in
onto another by maintaining proper tension and melding them together, Nevada-based Aerion, which is developing an advanced supersonic
reducing the scrap to zero. business jet. 8 The 12-passenger AS2 will be built mostly from carbon
fiber composite and is designed to fly over ocean routes at speeds up to
But his employer refused to invest in his idea. Not willing to take no for
Mach 1.4, or approximately 1,000 miles per hour. For MTorres, it turns
an answer, he founded his own company, MTorres in Pamplona, Spain,
out the same technological insights required to maintain tension in the
the city famous for the annual running of the bulls. Manuel’s young
splicing of paper also apply for more complex materials like carbon fiber.
company soon became a leader in paper converting.
Suddenly Boeing, Airbus, Fokker, Embraer, and many other aerospace
At the very beginning, MTorres started selling its splicers to S&S manufacturers needed the company’s help. Today, there is hardly a jet in
Corrugated Paper Machinery, an American company that built the air that has not been touched by MTorres.
corrugated paper lines. Over time, major brands like Weyerhaeuser,
The company now holds 140 patents and has plants in 10 locations across
Georgia Pacific, and Kimberley Clark became customers—each eager
four continents, including one just down the road from Boeing outside
for technological innovation. A decade later, the aerospace industry
Seattle. The Spanish manufacturing equipment and automation specialist
came knocking when it needed an innovative way to weave composite
is now a cherished partner throughout both aerospace and paper.
materials, like reels of carbon fiber, together for airplane wings, fuselages,
But the march of progress has hardly slowed. Like so many of the
manufacturers we spoke with, MTorres began to realize that not only did
it have a competitive advantage due to its experience and skills, it also
had a comparative advantage because of its data.

That data coupled with an algorithm and compute power—a combination

central to advanced machine learning and AI—is leading MTorres to
breakthroughs never before imagined. The ability to flawlessly laminate
carbon fiber onto an airplane wing at 60 meters per minute means not
only greater efficiency and productivity but also greater accuracy and
fewer mistakes. With AI, the company can also inspect its work in real
time. Never satisfied, they plan to increase that speed through AI-based
real-time inspection.

29 30
MTorres now is using advanced machine learning and AI in its
Automatic Fibre Placement machines to optimize for scrap reduction.
This is achieved via a drill that uses visual recognition to perfect tens of
thousands of holes per part, and an automated inspection module that
scans the surface of carbon fiber in real time looking for defects with
a high resolution camera and a laser. The defect analysis is done using
deep learning algorithms in the preprocessing and classification steps. In
aerospace, as the saying goes, failure is never an option.

Swarm Mentality

In locations around the world, Toyota Material Handling—the division

of Toyota Industries Corporation focused on logistics—is using AI to
reinvent the way it does business.

Axel Wahle, a genial leader at Toyota Material Handling, has spent more
than 30 years in the business. Rising through Germany’s much-respected
apprenticeship programs for both engineering and commerce, it’s Axel’s
job to bring the customer voice into everything the company does.

Toyota Material Handling is like many businesses around the globe

that are confronting the need to store, move, and deliver components
and finished product to help meet consumer demand. Their enormous
warehouses are fields of shelves, stacked floor to ceiling and loaded
with scores of car parts in every shape and size. It’s the job of people
like Axel to help manage the material flow for every SKU variation
using autonomous pallet drones and forklifts that snake their way
across warehouses to retrieve goods needed—whether for assembly in
manufacturing applications, or for consignments within the retail sector.

Not long ago, Axel presented a new idea involving AI that would
improve productivity and accuracy. It’s called “swarm” and represents
a revolutionary new approach to what the company calls “automated
handling.” Small robots swarm through the facility and use AI with
visual recognition to navigate. They carry loads to the correct location for
consolidation and onward transportation.

31 32
Toyota Material Handing uses Microsoft’s AirSim, which creates realistic TURNING INTELLIGENCE INTO ACTION:
digital twin environments to replicate vehicle dynamics and test how CREATING A DATA SUPPLY CHAIN THE FUTURE

autonomous vehicles using AI can operate safely in the open world. OF AI AND
MANUFACTURING

Further, the company uses Microsoft Bonsai, a novel approach using It’s AI, but Not as You Know It
machine teaching that analyses the images taken of an environment to In Detroit—a city famed for its role in the Second Industrial Revolution—
identify components such as racks, pallets, and other infrastructure. we visit a customer helping to define the Fourth Industrial Revolution.
Then the autonomous vehicle training takes place inside the AirSim
simulated environment. Auto parts giant ZF Group is pioneering data algorithms and technology
to make their production line more reliable and sustainable. In
The end vision is that Toyota Material Handing’s AI system can examine other words, detecting and predicting mechanical failures on the
the image of a shelf full of products, identify bar codes, and get the right manufacturing line and conserving energy. But to create the right culture
product to the right place. The vehicle is constantly learning and then of transformation, Georg Gabelmann, ZF’s Data Science Manager, says,
sharing what it’s learning with other vehicles. “Don’t call it AI. For me, AI is a buzzword.”
Axel said, “The swarm is for coordinated distribution, lean logistics, and
continuous flow—the right truck, the right task, on time, every time by A company needs to continue its path to
utilizing millions of calculations per second.”
capture quality data and to develop the
A human is ultimately responsible for validating that all the parts are
correct. Given the diversity of SKUs in which variations are so slight, human skills to improve machine learning
Toyota Material Handling entrusts final approval to human eyes and and eventually build true AI.
experience. The approach made sense to engineers, but the company
insisted that customers be consulted for feedback. More than 1,800 Georg is an industrial engineer with experience at SAP and thyssenkrupp
customers provided feedback on the swarm approach and, after careful who now leads the IT Innovation Team at ZF where they invent,
examination, Toyota Material Handling agreed that they should pursue showcase, and prove a range of new methodologies. Georg says, “True AI
it because the ROI for humans is so great and the accuracy is so high for is miscommunicated and misunderstood. Applied machine learning is AI,
the end consumer. and data science is the key. If someone came to me and said we are doing
AI, I would disagree. Our first step is machine learning.” Georg believes
What the two examples—MTorres and Toyota Material Handing— the company needs to continue its path to capture quality data and to
demonstrate is thanks to the digitization of assets and the infusion of develop the human skills to improve machine learning and eventually
AI to automate processes, these organizations are applying AI not only build true AI. And they are making progress. For example, a situation
to their current processes, but they are creating additional value and arose where a crucial tool on one of ZF’s lines was continually breaking
carving out competitive advantage. and no one could figure out why. Working with Microsoft, they trained a
The next customer stories will share how manufacturers are thinking model to diagnose the problem and, over time, by using more and more
about data in the same way as they consider other critical parts of their sensor and maintenance data, they improved the model and eventually
physical supply chain. were able to better predict the expensive tool failures.

33 34
Likewise, they’ve trained their data models to examine energy Milking the AI Potential
consumption. The company pays a premium for power when it surpasses
While cars might be driving themselves, dairy cows can’t milk THE FUTURE

a certain level. In the past ZF had to fly blind, but today the company OF AI AND
MANUFACTURING
themselves, but AI is helping one of our customers take a
uses an intelligent assistant based on reinforcement learning to better
step closer. The urgent phone call that a food manufacturer makes
predict energy consumption. An area of machine learning, reinforcement
four seconds after a failure on the dairy line is very different than the
learning is defined as being “concerned with how software agents ought
preventive call that a manufacturer receives four weeks before a failure.
to take actions in an environment so as to maximize some notion of
cumulative reward.” Georg explains that his machine learning system can The call four seconds after a failure is often addled by panic because
recommend when to shut down an air conditioning unit or to postpone a 50,000 liters of milk are at risk. The call after the prediction of a failure,
test run until later. four weeks before it happens, is somewhat less frantic.

Georg’s colleague, Robert Copelan, is based in Atlanta and has spent 33 Let’s meet someone who is trying to avoid the panic calls by using AI to
years in the automotive industry. As he listens to Georg talk about where makebetter predictions in the food manufacturing sector.
ZF is today, he cautions that the company is still very much in a proof-of-
Johan Nilsson oversees digital for Tetra Pak, a leading liquid food
concept phase. Progress is not yet enterprise-wide. Asked if he had advice
packaging manufacturer based in Sweden. Tetra Pak does aseptic
for other manufacturers, Robert replied, “Think big but start small. You
packaging and processing for major food and beverage brands like
have to think about what it means to have it adopted across the entire
Nestle, Tropicana, Coke, and Pepsi. Nilsson tells the story of how AI is
enterprise. Showcase and create awareness that can prove the value and
helping Tetra Pak prevent failures. Not long ago, a major dairy producer
get people on board.”
located just outside of Buenos Aires, Argentina, received a call predicting
a future failure. By connecting packaging lines to the Microsoft Azure
“Think big but start small.” Cloud, Tetra Pak can collect operational data from the dairy to help
He pointed out that some workers will fear machine learning and AI. ZF predict maintenance timing. The consequences of not predicting a
has worker councils designed to explain these advanced technologies.“AI failure are far-reaching as Argentina has strong import restrictions, so
will work in conjunction with us, getting rid of repetitive work,” Robert getting a replacement part from nearby Brazil or far-away United States
noted. “We strongly recommend not shutting down new tech initiatives,
instead trying to understand and take away the fear.”

Workers are already embracing the idea that having algorithms

with them when they need them is important. ZF’s “XReality” is a
combination of augmented and mixed reality in which wearable devices
like Microsoft’s HoloLens can help workers identify machines and parts,
access data, and help a worker or floor supervisor solve a problem.

ZF is helping to make workers more autonomous even as they help car

manufacturers build their autonomous and electric vehicles of the future.

35 36
 can take 96 hours to clear customs. Since a typical cow can produce 24
liters of milk every day, the loss of milk per cow, multiplied by a herd of THE FUTURE

hundreds, can be the difference between profit and loss. You can’t OF AI AND
MANUFACTURING

simply switch the cow on and off during a manufacturing failure.

“We can plan with parts so there is no fuss. No loss of production time,”
Johan boasts.

For many customers, Tetra Pak is the largest supplier they have. They
rely on Tetra Pak every day of every month. Having a high-quality
relationship is essential. Tetra Pak produces 600 data points per piece
of equipment with high frequency—too much data for any human to
analyze in real time. The company increasingly is a user of every Azure
feature, and engineers are testing blockchain on the Microsoft cloud,
which is tamper-proof.

Liquid food factories are held to very strict quality management

standards to secure food safety. Those not relying on the cloud and AI
cognitive services might be using paper or multiple spreadsheets, which is
labor-intensive and creates risk.

The Connected Welder

Predictive maintenance is driving machine learning and AI adoption

throughout the manufacturing sector, but so too are operational
efficiencies and improving productivity.

Colfax, based in Annapolis Junction, Maryland, provides a case in point.

The company, known for its expertise in welding and pumps, has grown
from a specialty fluid-handling company to a multi-platform diversified
industrial company serving hundreds of global industries.

Colfax is several years into its “Data Driven Advantage” digital strategy,
in which it is deploying a global common footprint that will enable the
company and its subsidiaries to take advantage of data no matter where
it is generated. Ryan Cahalane, a controls engineer turned technology
executive with a long resumé in manufacturing, is helping to lead the
transformation by aligning Colfax’s systems across all infrastructures.

37 38
“We used to be the nerds at the dance no one wanted to talk with, but “There’s a huge opportunity in the ‘brownfield’ (existing/legacy).
now we’re the most interesting people at the dance,” he laughs. Where it can have a meaningful impact now, right away, is unlocking THE FUTURE

value from existing assets. And empowering people on the frontline OF AI AND
MANUFACTURING
One example is Colfax’s ESAB subsidiary, a leading global provider of
versus just being tools to help data scientists back at corporate.” For the
fabrication solutions. With men and women operating their welding,
near term, AI is really about ‘augmented intelligence,’ helping the human,
cutting, and gas management equipment in almost every industry and
more than artificial intelligence.
application imaginable, it was very difficult to be able to identify when
customers were running into issues, determine the patterns leading up to
the event, or make recommendations. In response to the challenge, ESAB
“AI is really about ‘augmented intelligence,’
had the ingenious idea of creating “the connected welder.” By connecting helping the human, more than artificial
every welder to the platform, AI is lighting up the ability to provide
valuable insights into things like job quality, efficient use of material, intelligence.”
optimizing consumables, and even alerting users to potential supply
But most plants have a mix of new and older equipment, and often the
chain issues. As a result, ESAB can improve job costing and quality
expertise needed to operate at full potential is held by just a few people.
for their customers, or help them make better decisions about critical
“People can make or break the quality of the work on a given day,”
production resources and equipment utilization. With all that data, and
Ryan points out. “An operator of a line or supervisor is a bit like the
with the ability to finally “see” trends in the huge diversity of customer
conductor of a symphony, and their experience and mindset can make
applications, AI will in turn help make recommendations back to ESAB
the performance just average or really amazing. Many can tell just by
on how to improve designs and their own operations.
the background sound of the operation how things are running, where
“AI can find patterns we didn’t know before—weaving together a fabric problems are beginning, and how long they have before it really affects
that extends across the entire value chain,” Ryan explains. “As it unfolds, overall performance.”
we will get insights at lots of different levels. We can then make better
With the changing workforce in manufacturing, including the retiring
hypothesis and correlations, at a rate of innovation that was never
of experienced baby boomers, much of this experience and capability
achievable before.”
is walking out the door. While more automation and robotics may
Ryan has spent his life in and around manufacturing. Originally from help, humans are still critical in most operations. This means that
Akron, Ohio, his father was a manager at Goodyear, which meant living manufacturers need every operator to perform like a concert master.
wherever Goodyear had its tire and rubber plants, be they in France,
Just like the background noise example, AI can help identify patterns
the Congo, or Morocco. After studies at Purdue and Case Western
that less experienced operators might miss, accelerate learning new skills
universities, Ryan joined Goodyear himself, but he eventually went to
or amplify skills operators already have, and help operators connect
Deloitte because he wanted to see the bigger picture in his chosen field.
with other experts to collaborate on a particular situation. The old way
Ryan recognizes manufacturing still has a long way to go, but he also is of controlling an operation was static, often not taking advantage of the
excited by the meaningful and accelerating progress that lies ahead. incredible capacity operations teams must have to improve performance
or innovate.
“But it isn’t all about new products or ‘greenfield’ operations,” he says.

39 40
Now, through AI, the operator is empowered with more knowledge
about the baseline throughput and quality, and now is provided with
recommendations in real time on how to improve performance. In
this case, as with so many others that Colfax have found, operators
can now use their full potential to evaluate and execute the best
possible performance.

As Ryan observes, “People don’t wake up and think about digital

transformation. In fact, in the real world of production operations, it
makes them recoil. If not positioned and focused properly, AI could cause
a real rift with people if we’re not careful. Augmented intelligence is a
more powerful, and timely, message. That is something people can get
behind. This will make me a better worker and compete more with the
offshore, low-cost worker that threatens to take my job.”

From car parts to beverage packaging, from industrial welding to

elevators, this diverse range of manufacturers demonstrates the reach
and impact of AI. These customers understand the value of data and
how essential it is to making their products safer and more reliable. Each
manufacturer recognizes their critical role in a larger supply chain and
that AI will be an ever-increasing asset to liberate and harness data
across that supply chain to meet their needs and the needs of
their customers.

Where to from Here?

In story after story, we met individuals who care deeply about the future
of their business as well as the future of their workers. We heard from
managers and plant workers who understood that digital and cultural
transformation work hand in hand. We also witnessed what happens
when you remove the barriers around an organization’s data estate and
convert information into innovation.

And while these stories show us what realizing tech intensity is all about,
they also remind us that people are at the heart of this transformation.
In the next chapter we look carefully at questions and issues surrounding
the human impact and the implications for tomorrow’s AI workforce.

41 42
AI AND MANUFACTURING: The goal was to rush data-based insights to people on the
MICROSOFT’S CULTURAL TRANSFORMATION plant floor.

They lit up Microsoft’s Power BI business analytics to produce data

visualizations and dashboards that ensured plant managers and
The time difference between America’s Pacific Northwest and supervisors alike were seeing the same information at the same
Shanghai, China, is 14 hours. While senior executives overseeing time—one source of truth. The dashboards funnel all the telemetry
the manufacture of Xbox consoles, Surface computers, and other data that’s collected, such as cycle rate and the pressure of a
Microsoft hardware products get their day started, it’s the wee particular press, into a model to detect shifts and drifts so they can
hours of the following morning 5,700 miles away. The need to respond before a part fails or a line has to shut down. While some
view real-time information and insights is crucial, for both the plants do this for a single machine, Microsoft created a system that
headquarters in Redmond, Washington, and plant managers in looks at all machines at once. Finally, the team also created AI for
Shanghai. planning purposes. Using a range of statistics and information from
the sales team, inventory, promotions, and other data, the team also
When Mark Klinkenberg left Intel to join Microsoft in 2008, the
can have more accurate demand forecasting.
software company was preparing to ramp up its hardware business,
which meant ramping up its manufacturing know-how. Back then, Mark laughs when he retells the story that frontline factory workers
data from the assembly line sauntered into various databases and started to ask if he could pause data before it was sent to executives.
spreadsheets. Analysts crunched the numbers and produced reports They worried about getting yelled at prematurely, but executives
for a conference call the following day with plant and executive learned quickly that everyone was acting urgently on the same data.
management. The data was old and not always actionable. Mark
They’ve been able to move from 40 percent orders committed within
and his team began to ask, “How can we get smarter faster?” The
five days to 95 percent committed within 48 hours. There’s been a
team needed to look for patterns, relationships, and causality across
$50 million reduction in errors and omissions year over year, and
the plant with greater speed, accuracy, and efficiency in order to
they’ve saved another $10 million in just one year due to scrap
develop better solutions for building products and improving yields.
reduction, yield improvements, and process optimizations. Demand
“We had a lot of information,” Mark recalls. “But how could forecasts are 15 percent more accurate. But one of the greatest
we get smarter?” benefits has been a more engaged workforce from top to bottom.

AI became a critical tool for Microsoft’s manufacturing team in Mark and his colleague Darren Coil warn that the cultural
China and the United States. Together the team executed on three transformation needed for AI to grab hold in manufacturing cannot
strategic fronts: get connected, become predictive, and grow to be be underestimated. In fact, they note, the culture will fail before
cognitive. In other words, they needed to connect as many devices, the technology does. The cultural change won’t happen through a
instruments, tools, and people as possible, build algorithms that webinar or reading a manual. It must be a combination of top-down
could convert the massive amounts of data this connectivity enables and bottom-up change. In their case, the corporate vice president in
into predictive insights, and ultimately train the model to be charge decided to lead by example—executive role modeling.
cognizant—to think—of solutions to augment human capabilities.
43 44
No more static data presented in Excel or PowerPoint. All
presentations and all meetings took place gathered around live data
and analysis using Power BI. Whether you were a factory worker in
China or a business manager at headquarters in Redmond, you had
to use Power BI.

“Show them, don’t tell them,” Darren recalls. “Reinforce the lessons
and keep coming back. Tell what you’re going to tell them, tell them,
and tell them what you told them—repetition.” 

What Darren and Mark have observed is that executives need to

set expectations and role model as a first step. It’s also critical to
involve workers in the process so they can see how the machine
learning and AI algorithms are trained. This builds trust and
buy-in. Helping them understand these systems will give you
better information to make better decisions. Then, the executive
needs to step away and come back a few months later to see if the
transformation is taking hold. In Darren and Mark’s example, the
executive returned to find that the factory team was using and
implementing data and AI to solve problems and have quicker
impact. In fact, they started to become keen consumers of data,
demanding access to more data and information—a flywheel or
virtuous cycle begins to gain momentum.

A cautionary tale: Mark and Darren point out that one obstacle
to cultural transformation is that workers are accustomed to
having their performance evaluated using the earlier system. A
plant manager might say, “The old system showed me running
at 95 percent yield. That was my goal and I was rewarded
accordingly. The new system and data show me at 92 percent.” Clear
communication during the transformation is important. Machine
learning and AI data models will be more precise, and senior
leaders and workers must be understanding and flexible during
this change.

45 46
CHAPTER 2

AI AND THE
MANUFACTURING
WORKFORCE

47 48
AI AND THE MANUFACTURING WORKFORCE companies to follow principles such as demanding that AI systems be
transparent, be built with an ethical black box, adopt a human-in- AI AND THE

command approach, and ensure genderless and unbiased AI.12 MANUFACTURING

WORKFORCE

Meanwhile, the Organisation for Economic Co-operation and

Development (OECD), a group of 36 countries accounting for 80 percent
On a frightfully cold January day in 2019, just over a year before the U.S.
of world trade, recently observed in a set of recommendations that AI
presidential primary, New Hampshire voters were treated to a parade
may have disparate effects within and between societies, including the
of potential presidential candidates making their case. Headlines about
deepening of income inequality and widening of the skills gaps.13
their auditions appeared in Twitter feeds next to President Trump’s
excitement that “last year was the best year for American manufacturing In this chapter we explore the implications of factors ranging from
job growth since 1997, or 21 years … and it’s only getting better.” economic dislocation and workplace safety to the future of work as they
relate to manufacturers and policymakers alike. Just as we did in the
But newspapers were also reporting that manufacturers might be cutting
previous chapter, we begin with the voices of our customers and partners
back. The Wall Street Journal reported that technological change
and those on the frontlines.
sweeping the auto manufacturing industry is forcing job cuts.9

The previous summer, a union representing culinary workers in Las A NEW RELATIONSHIP WITH TECHNOLOGY: CREATING
Vegas, Nevada, issued a press release noting that it would be negotiating
A SAFER AND MORE SATISFYING WORKPLACE
“language regarding increased technology and the effects automation has All the manufacturers we interviewed agreed that worker safety must
on jobs and workers.” A Huffington Post article reported that “workplace continue to be the highest priority. After all, much of the momentum
experts say unions need to figure out how to help workplaces and workers behind robots is driven by a desire to replace human laborers doing
adapt to new technologies to reduce layoffs if workers are to have hope of repetitive and often dangerous tasks.
surviving and even thriving in the face of this threat.”10
But even with these advances, manufacturing remains prone to safety
At UNI Global Union, an organization representing 20 million workers challenges. For example, in the United States, muscle sprains, strains,
from more than 150 countries, the issue of ethical AI and the implications and tears are the leading type of injury in manufacturing, resulting in
for the workforce is something they have been very active and vocal on. an average of 10 days away from the workplace for those suffering from
In 2016, they called for the creation of a global convention on the “ethical these injuries.14 In Europe, material-handling equipment like forklifts can
use, development and deployment of artificial intelligence, algorithms cause accidents, and accidents cost lives and money—€5 billion per year
and big data.”11 according to one estimate. As one customer told us: “It can be unsafe
More recently they issued 10 principles for ethical AI as it relates to the today, which is why we are so keen to move to autonomous systems. Our
future world of work. According to their report, while some workers top priority is to get to zero accidents.” Another customer says he wants
are already losing their jobs, they also state that “AI, machine learning, AI that can improve safety monitoring and control in real-time fashion,
robotics and automated systems can also benefit workers.” UNI Global to make sure equipment is being run as intended. He wants cutting
Union called on companies to work with labor and encouraged those machines that are more sensitive and forklifts that know when people
are nearby.
49 50
It would be hard to find someone more steeped in the field than Bazmi “The only thing I fear about AI is that
Husain, a 38-year veteran of Swedish-Swiss technology giant
people will fear AI.”
AI AND THE

ABB, which specializes in robotics and discrete and industry MANUFACTURING

WORKFORCE

automation technologies. Listening to Bazmi is rather like attending a

ABB’s automated control systems react to planned events and are
TED talk. The future he envisions is one where AI and robotics augment
there to help people be more efficient and productive. But as with all
human capabilities to free us from dangerous occupations and repetitive
programmable systems, there comes a point where it no longer pays to
tasks while delivering unprecedented improvements in living standards.
automate—where conditions become too complex or unpredictable to
“The only thing I fear about AI is that people will fear AI,” he remarks, make programming feasible or worthwhile—and humans have to step in.
clearly concerned that trying to put the brakes on AI would be wasting
“With AI, we can address this challenge,” Bazmi explains. “Rather than
an opportunity to improve people’s lives.
blindly following rules, AI is able to learn from prior experience and
ABB pioneered the modern robotics revolution in 1974 with the first human intervention and so react to unforeseen situations in the most
microprocessor-controlled industrial robot, and Bazmi has been part appropriate way. Not only does AI take the pressure off people—reducing
of the evolution of smart robotic solutions since he joined the company the potential for human error—it augments human capabilities by
in 1981. Today, as ABB’s Chief Technology Officer, he is focused on allowing people to focus on tasks that humans are ideally suited to doing,
innovation, with particular emphasis on AI for industrial applications. supported by autonomous systems.”

“Robots used to be unaware of their surroundings, so we had to put them ABB is working on autonomous systems for a variety of industries.
in cages or behind fences to keep humans out of harm’s way,” Bazmi In December 2018, the company achieved a breakthrough toward
says. “But with advances in sensor technology and industrial AI, robots autonomous shipping when ABB’s new intelligent autopilot enabled a
are now able to sense their environment—to ‘know’ what is going on ferry captain to remotely pilot a passenger ferry through a test area in
around them—which means we have been able to free them from their Helsinki harbor.
cages to collaborate with humans. The result has been safer working
With AI, ABB can better predict when a machine, robot, or system is at
environments and far higher productivity.”
risk of malfunction. For human operators, knowing when to intervene
preemptively avoids costly downtime and unpleasant surprises. To
appreciate just how unpleasant an unexpected breakdown can be,
imagine what happens when an offshore wind turbine suddenly stops
working. A standard wind turbine is a 116-foot blade atop a 212-foot
tower with a total height of 328 feet—30 or more stories in the air—and
wind turbine speeds reach 200 mph. Without AI, operators have no idea
when they might have to send a crew out to sea to repair the turbine. In
bad weather, it might be days before action can be taken.

51 52
With sensors, data, and machine learning algorithms, ABB can not only
predict with 95 percent accuracy when a turbine is likely to fail, but the
company can provide prescriptive information on remaining useful life
under different operating conditions. Armed with this information,
an operator can schedule preemptive maintenance at the most
opportune moment.

With AI to augment human capabilities, ABB can make its industry safer,
cleaner, and more productive than ever before to the benefit everyone.

THE TALENT SUPPLY CHAIN:

A SHARED AGENDA FOR SKILLS AND EMPLOYABILITY

The debate around AI-induced automation and the workforce is

extremely vivid, but it’s rarely based on strong empirical evidence,
according to a recent LinkedIn study.

LinkedIn, a Microsoft subsidiary, is a professional social network with

nearly 600 million members worldwide. LinkedIn creates what they call
“The Economic Graph,” a digital representation of the global economy
based on data about those members, 30 million companies, 84 thousand
schools and universities correlated with 50,000 skills and 20 million
open jobs.15

LinkedIn’s research into emerging skills around the world has shed light
on a few growing trends. First, AI skills are among the fastest-growing
skills on LinkedIn, and saw a 190% increase from 2015 to 2017. Second,
industries with more AI skills present among their workforce are also the
fastest-changing industries. Finally, in the same time period, when all the
different skills held by an industry’s workforce are compared, it’s clear
which industries have changed the most by how much their overall skills
makeup has changed.

The manufacturing sector ranks fifth behind software and IT services,

education, hardware and networking and finance for having the most
need for AI talent.

53 54
 In the United States, Jay Timmons, the President and CEO of the
National Association of Manufacturers (NAM), has thought a lot AI AND THE

about the future of the workforce and the implications for the MANUFACTURING
WORKFORCE

constituency he represents. His conclusion?

“Our shared belief [is] that every one of us can contribute to the success
of our companies, our communities, and our country. Manufacturing is
more than just technology and machines; our industry is about people
and the potential we can unleash.”

According to a NAM survey, 82 percent of manufacturing executives

indicate they believe the skills gap will impact their ability to meet
customer demand. 78 percent believe the skills gap will impact their
ability to implement new technologies and increase productivity.17

As a result, NAM has explored a range of policy recommendations,

including the need to adapt to the changing needs of the modern
manufacturing workforce and the changing attitudes individuals have
toward their work. The organization recommends building a
system that delivers a sustainable pipeline of strong, mid-skilled
manufacturing talent.

The question of how best to prepare workers for the future occupies
labor and trade unions everywhere, including Scott Paul, President of
the Alliance for American Manufacturing (AAM). AAM is a partnership
established in 2007 by some of America’s leading manufacturers and the
United Steelworkers union.

Scott notes that conversations about workforce reskilling or upskilling

tend toward two camps: the “philosophical salons of influential coastal
cities” and the more practical factory floor discussions in manufacturing
This is important data as it demonstrates that despite the huge demand
towns. Both are valuable, but increasingly we need to get
for AI skills in manufacturing, the talent supply is not keeping up.
more practical.
Researchers at LinkedIn also point out that governments are currently
It’s a challenge Microsoft’s Portia Wu thinks a great deal about.
devising their own AI strategy, craving for more objective data and
insights about AI progress and hindrance.16

55 56
Before joining Microsoft as Managing Director of U.S. Policy, Portia spent Second, these insights need to be shared with policymakers,
her career developing and implementing labor and workforce policy, educational institutions, and, most importantly, workers so that AI AND THE

including serving as Assistant Secretary for Employment and Training at everyone is operating with the same information. MANUFACTURING
WORKFORCE

the U.S. Department of Labor.

Next, education and training opportunities must be refocused so they
“Within manufacturing, we’re in a middle chapter, not at the beginning,” can meet those skills needs. This means not only degree programs, but
she explains. “The industry has undergone a lot of change already from also certificate programs, online learning courses, and internships and
robotics and automation.” apprenticeships. All of these must align with these skills and jobs.

Finally, these efforts need to be evaluated in real time in order to share

“Within manufacturing, we’re in a middle results and adjust as necessary. The goal should be helping to transform
chapter, not at the beginning,” she people and their ability to succeed in the jobs of the future as those jobs
are created.
explains. “The industry has undergone
Across town from Portia’s office in Washington, D.C., is the Siemens
a lot of change already from robotics Foundation, part of the Global Alliance of Siemens Foundations, whose
mission is to ignite and sustain today’s STEM workforce and tomorrow’s
and automation.” scientists and engineers.

When studying the issue of labor displacement and new skill The Siemens Foundation focuses its resources on a particular theory of
development caused by technological advances, she notes that change: middle skill development for young adults aged 16-30.
policymakers need to clearly identify which workers are being discussed.
Crystal Bridgeman is the Senior Director for Workforce Strategies at the
For example, needs may be very different for workers who are being
foundation and cites OECD data indicating that Americans rank lower
displaced, for workers whom companies are helping to gain added
than citizens of other countries on numeracy and problem solving, which
skills, or “upskill,” and for workers who program or design AI systems.
limits their job opportunities in STEM occupations like manufacturing,
There are other important dimensions to consider, including whether
healthcare, IT, and energy. Often what is needed is more than a high
workers are just entering the field, or if they are mid-career or late career.
school degree but less than a four-year bachelor’s degree.
Depending on the circumstance, policy solutions may differ. To improve
our training pipeline, Portia suggests some key steps are needed: The Siemens Foundation has set three objectives for its funding
and advocacy:
First, business and industry need to define the needed skill sets that map
to the jobs and careers of the future. Where is the demand, and how can First, it wants to be an accelerator of training models that are backed by
we standardize job titles and descriptions? In the past, the profile of a evidence or show early promise. In particular, this means work-based
welder was very clear; but, as the pace of technology change continues learning and apprenticeships. The foundation has partnered with the
to accelerate, the role of a data labeler or data analyst may be very National Governors Association (NGA) in 17 states and with the U.S.
different from company to company. Moreover, new jobs and roles will be Department of Labor through the American Apprenticeship Partnership,
emerging every day. among others.

57 58
 Second, their effort is designed to support training and development
initiatives that emphasize high-quality, digital skills and that blend AI AND THE

both classroom study and hands-on application in the workplace. MANUFACTURING

WORKFORCE

Finally, from an advocacy perspective, the foundation is working to

change American perceptions from a “college or nothing” mindset to
one that embraces a variety of educational pathways that lead to
satisfying careers.

Apprenticeship programs in Europe, especially in Germany and

Switzerland, serve as an inspiration for a more robust post-secondary
system. There, the idea of an apprenticeship is not a point of contention,
Crystal observes. Government, business, and industry are all on board.
Apprenticeships are transferrable to higher education, and higher
education is transferrable to apprenticeships. To truly emulate the
European model will be difficult in the United States, where
education policy is set much more at a state and local level than at
the national level.

“We need an ecosystem that brings together the right partners and high
standards that are measured against valid benchmarks—industry and
academic standards,” she said.

“We need an ecosystem that brings

together the right partners and high
standards that are measured against valid
benchmarks—industry and academic
standards,” she said.
Across the Atlantic, the experience with apprenticeship is very
encouraging. Unemployment after apprenticeship is very low in Europe
compared with unemployment or underemployment following university
graduation in the United States.

59 60
This is the assessment from Michel Servoz, Senior Adviser on AI and from optimization. Will they reinvest in the workforce? Will they look
Robotics to European Commission President Juncker. Michel notes that to the future and make plans to compete in new markets? Or will they AI AND THE

an apprenticeship not only develops skills but is also a great introduction simply take the profits? Workers and their unions worry that they will MANUFACTURING
WORKFORCE

to the world of work, including the necessary relationship-building and not be part of the conversation as advances are being considered
communications skills that are needed. and deployed.

His view is that rethinking education is job number one for policymakers Brad Markell from AFL-CIO Industrial Union Council—who is also
everywhere. He sees a skills mismatch—both technical and soft skills— a long-time member of the International Union, United Automobile,
that if not resolved will lead to grave economic consequences. Aerospace, and Agricultural Implement Workers of America (UAW)—
calls this “procedural justice,” the notion that you must involve workers
Not only do students need technical and digital skills, but they also need
in the creation of processes and technologies in order for those processes
to develop the skills that are not likely to be automated anytime soon—
or technologies be truly helpful to those workers. No one, he notes, wants
creativity, problem-solving in an unpredictable environment,
to end up with “bargained acquiescence,” which Black’s Law Dictionary
and empathy.
defines as “consenting without any enthusiasm.”
Moreover, education must be lifelong. The current model of education,
“You want workers who say, ‘I like my job,’” Brad advises. “You want
career, and retirement no longer works. Training and education will
workers who say, ‘I am challenged rather than just doing routine things
become a lifelong pursuit.
over and over.’”
Those countries where this is happening are seeing the rewards.
Brad points to General Motors and Ford as examples of companies that
Jobs that once left for low-wage countries are beginning to return. Adidas are investing in training and smart factory simulation to help bring
closed plants in Malaysia and Vietnam and opened plants in Germany workers along. For example, 3D printing is being experimented with for
and the United States. Philips did the same, returning jobs outsourced to toolmaking and production of small fixtures for autos.
Asia back to The Netherlands.
“It’s not a peanut butter approach,” Brad said. “It’s application
“There is huge anxiety about the future of work among trade unions,” by application.”
Michel said. “Productivity and wages are stagnant, so they are
Another green shoot—a harbinger of things to come—is a promising lab
rightly concerned.”
at the Digital Manufacturing and Design Innovation Institute
And while these concerns about the impact of technology are universal, (DMDII) in Chicago.
Scott Paul, from AAM, says, “Tech is a fact of life. We are not luddites.
We’re not standing in the way. But there is also a realization that “You want workers who say, ‘I like my job,’”
it requires a different set of skills. It’s simplistic to say AI will kill
manufacturing jobs, but it will transform them.” Brad advises. “You want workers who say,
While there is agreement on its inevitability and benefits, workers are ‘I am challenged’ rather than just doing
concerned about what companies will do with the savings that result
routine things over and over.’”
61 62
The lab is a public-private partnership in which workers learn new But we also need to be concerned. Concerned that we have the right
skills inside a simulated smart factory. While still small scale, the lab training and education infrastructure to create a future workforce that AI AND THE

demonstrates what’s possible when workers learn new digital processes, reflects the future workplace and a social safety net that protects all MANUFACTURING
WORKFORCE

analytical tools, quality assurance and monitoring, component workers in our modern economy.
installation, and how to collaborate with digital and AI engineers to
If every manufacturer will be using AI as part of its operations, then the
improve systems.
talent needed will be different in the future than it is today. We need to
These workers emerge with heightened numeracy and digital skills that develop new approaches to training and education that enable people to
are transferable to other factories of the future. acquire the skills that employers will need as technology advances and
create innovative ways to connect workers with job opportunities.
Alongside the huge opportunities that exist to upskill the exiting
manufacturing workforce, we may also need to look at new ways to As we move forward, it will be essential for a new coalition of interests to
support people for whom disruption means switching to a different career come together to explore how to best support workers, create economic
or spending some time out of the workforce. We’ll need to adapt new opportunity, and plan for the future. But this coalition of interests needs
labor laws and policies, many of which were built in the age of the steam to do more than just focus on the workforce, as critical as that is. This
engine, not the search engine. coalition needs to help shape the broader ethical and policy frameworks
that will foster responsible innovation. The next chapter looks at how we
Now a century later, these labor laws and policies are no longer suited to
might achieve this.
the needs of workers or employers. Health insurance and other benefits,
for example, were designed for full-time employees who remain with a
single employer for many years. But they aren’t as effective for individuals
who work for multiple companies simultaneously, such as those who work
in the gig economy—think Uber drivers or those who change employers
regularly. There is an increasingly pressing need to review these benefits
and ensure they provide adequate coverage for all workers in the digital
economy, as well as a sustainable contribution structure for business.

WORKING TOGETHER
So, what does all this tell us? Should we be excited or concerned?
Perhaps it tells us we need to be a little bit of both.

We should be excited for the opportunity that AI brings to help create

more compelling jobs in manufacturing, jobs that are working alongside
AI rather than being replaced by AI. That AI will reduce the number
of low-value, repetitive, and, in many cases, dangerous tasks. This will
provide the opportunities for millions of workers to do more productive
and satisfying work.
63 64
CLOSING THE AI SKILLS GAP: Infrastructure Management, as none of the other disciplines will
MICROSOFT’S COMMITMENT succeed without a robust, technical environment supporting Data
Management and Data Science and AI. Each of the disciplines has
its own track of learning curricula, and this is how we have designed
our online learning programs that span the entire journey from
Microsoft has a shared goal with our customers to identify and core foundational knowledge to specific technical skills needed in
develop the right set of skills they need to leverage the potential the workforce.
of AI in their business. We are doing this by implementing skills-
Armed with this knowledge from our customers and partners, we
building solutions across education systems, public agencies, and
created a suite of programs that are available worldwide to address
workforce industries to provide a breadth of skills training to meet
the skills needs:
the need of individual learners and support countries in building
their talent pipelines. So, how is Microsoft helping address the skills The Microsoft Professional Program (MPP): The MPP was created
gap and enhancing employability for our customers? to help people gain technical job-ready skills and get real-world
experience through online courses, hands-on labs, and expert
First, we started by interviewing and researching organizations
instruction, which culminates with a capstone project that
who successfully leveraged AI as well as those whose AI initiatives
allows learners to demonstrate their learning. Throughout the
were unsuccessful. Additionally, we spoke with executives familiar
program, students receive certificates that accrue to a Certificate
with the AI initiatives of companies that are considering the
of Completion for the entire MPP track upon completion of the
introduction of AI into their organizations. The organizations we
capstone. Each track is aligned with a tech job role in today’s
explored included companies in the finance, real estate, education,
workforce. Because MPP learners are required to demonstrate
manufacturing, healthcare, and retail sectors, including: Severstal,
proficiency in the skills they acquired, employers have the
Nissan, Goldman Sachs, Bloomberg, MI6, Hirotec, Arizona State
confidence they need to hire skilled individuals for in-demand
University Skysong Center for Innovation, University of Kansas
roles, which puts potential candidates onto the path to a rewarding
Medical Center, Rolls-Royce, Burberry, Walmart, and Tesco.
career. Among an extensive list of tracks, the MPP offer tracks in
​ ur work surfaced several key takeaways. Most fundamental is
O Data Science, Artificial Intelligence, AI Developer, and
agreement that there is a lack of AI-related skills in the marketplace Data Analyst.
and that demand for data and AI skills will continue to outstrip
Learning as a Service (LaaS): LAAS helps our customers with
supply. The research also revealed agreement that we are in the
Azure-based learning solutions to train their employees or students
early phases of this digital revolution and that the skills and
on the latest Microsoft technologies. This program combined
capabilities required today will be table stakes in the future.​
with MPP provides customers with differentiated and tailorable,
​ nother key takeaway from our research is that there are three
A technology-neutral skill building in high-demand job areas such as
core disciplines that must be present and active in every AI project AI Engineering, AI App Developer, and more.
to be successful: Data Science and AI, Data Management, and AI
Business Integration. We considered including a fourth discipline,

65 66
The Microsoft Imagine Academy (MSIA): Reaching over 8 million task that is bigger than any one company. That’s why Microsoft
students and educators annually as of 2019, MSIA provides the Philanthropies is working with businesses, governments,
industry-aligned curricula and certifications to build competencies nonprofits, and educational institutions to help those already in
and validate skills for high-demand technologies. MSIA offers the workforce, those trying to re-enter the workforce or build new
courses and certifications in four paths: Computer Science, IT skills, those who are seeking to gain work experience, as well as
Infrastructure, Data Science, and Productivity. equipping future students with the digital skills critical to ensure
both financial stability and opportunities for growth.
LinkedIn Learning: This platform combines an unmatched library
of more than 13,000 courses taught by real-world experts with Microsoft is leveraging its position as a leading global technology
LinkedIn data and insights drawn from over 575 million member company to develop technological solutions, and make strategic
profiles and billions of interactions, which gives employers a unique investments in skills development and employability programs, as
and real-time view of how jobs, industries, organizations, and skills we work to ensure that this industrial revolution and the technology
are evolving, while also helping leaders identify the skills their driving it, creates economic opportunity for all. Microsoft
organization needs to succeed. Philanthropies partners with nonprofits to increase equitable
access to high quality computer science education, specifically
AI Business School: This online course is a master class series
seeking to reach young people who are least likely to have access
that aims to empower business leaders to lead their organizations
to it. Around the world, more than 80% of the young people who
on a journey of AI transformation. Designed specifically from a
benefit from Microsoft Digital Skills grants come from underserved
nontechnical perspective and created in collaboration with INSEAD
communities, and more than half are female. These efforts support
(a graduate-level business school with campuses in Europe, Asia,
our commitment to ensure all people have the opportunity to learn
and the Middle East), the course materials include brief written case
the skills they need to succeed in an AI-enabled world.
studies and guides, plus videos of lectures, perspectives, and talks
that executives can access online. A series of short introductory
videos provide an overview of the AI technologies driving change
across industries, but the bulk of the content focuses on managing
the impact of AI on company strategy, culture, and responsibility.

Microsoft Philanthropies: By 2020, more than 800 million people

will need to learn new skills for their jobs and two-thirds of
students today will work in jobs that do not yet exist. Not only does
this skills gap impact prospects for individuals, it has a systemic
effect on the ability of companies, industries and communities to
realize the full potential of this digital transformation. Addressing
this issue involves changing the way people are educated and
trained, and the way companies hire and support their employees.
Reshaping the labor market for the 21st century is an enormous

67 68
CHAPTER 3

FOSTERING
RESPONSIBLE
INNOVATION

69 70
FOSTERING RESPONSIBLE INNOVATON Next to the farming demonstration is the actual cockpit of a Singapore
Airlines passenger jet using a Trent 1000 Rolls-Royce engine for the FOSTERING

Boeing 787. Singapore Airlines came to Microsoft asking for AI and RESPONSIBLE
INNOVATION

machine learning to accurately prescribe the fuel levels for each flight—
the optimal level that takes into consideration weight, wind, distance,
and engine efficiency. Rolls-Royce is extending its manufacturing
business into engines-as-a-service for prescriptive maintenance
and operation.

More than 25 sensors track engine health data, air traffic control, route
restrictions, and fuel usage to uncover data insights that will enable
airlines to improve operational performance and fuel efficiency.

In demonstration after demonstration, the conclusions are obvious:

Most days on the Microsoft campus outside Seattle, policymakers A thriving manufacturing sector and the use of the latest technologies are
mingle with manufacturers and representatives from other industries in inextricably linked, and that a thriving manufacturing sector is a priority
the company’s Executive Briefing Center, where demonstrations of AI, for the broader economy.
drones, robots, augmented reality, and other advanced technologies are
In fact, according to a report by the World Economic Forum (WEF) on
on display. What’s on display is not futuristic: It’s here today.
manufacturing and the future, manufacturing has the largest multiplier
One example is a Minnesota family’s corn farm. Neatly laid out on a 3D effect of any economic sector.18 The WEF report points out three
topographical map, the demonstration of the farm shows how weather, important areas for policymakers, business, and civil society alike:
seed data, satellite imagery, historical growth, soil moisture, and other
1. Public- and private-sector interests need to converge: This
variables inform farmers how to optimize corn production for companies
highlights the need for new industrial policy collaboration models,
like Land O’Lakes and Purina.
starting with new skills development models.
One of the Microsoft demonstration teams points out that some of the
2. A new manufacturing language must focus on capabilities and
most technologically advanced companies look on in disbelief, smiling
global value chains: To better anticipate underlying constraints
and asking themselves aloud, “How are we being beat on AI by a bunch
on industrialization, we need to move beyond the typical gross
of farmers?” But those in AI-informed farming respond that, at the end
domestic product measures.
of the day, it’s the farmer—not the machine—that is entrusted with the
ultimate decisions. 3. New, connected business industrial models: This can help
prepare for the potential Fourth Industrial Revolution.
In these examples, AI solutions help the farmer, first and foremost. It’s the
farmer, using machine learning to decide against a fungicide in preference While the WEF report also suggests these areas require further
for a John Deere planter that can put down variable seeds based on which discussion and that further research between the public and private
strain of seed works best in what soil. AI is the tool of the farmer. sectors is needed, it does pose some immediate questions: How can

71 72
further convergence of the public and private sectors, as well as For our part, we believe that AI will be the defining technology of our
civil society, be promoted? How can multiple policy initiatives and time. Like the discovery of electricity or the development of the steam FOSTERING

instruments be aligned in one direction (e.g., supplier development engine, AI will have the power to fundamentally change people’s lives, RESPONSIBLE
INNOVATION

programs, promotion of foreign direct investment, upgrading transforming industry and transforming society. It is therefore even
infrastructure, skills development)? At what pace have countries and more imperative that the development and deployment of this
businesses adapted to the changing manufacturing environment? How technology adopt clear principles to govern its creation and use.
should industrial policies be focused to match the development of future
In his book, Hit Refresh, Microsoft CEO Satya Nadella writes that just
capabilities? How can countries effectively move up the value chain?
as our ethics, values, and laws have been developed and evolved over
These questions are not easily answered, particularly given the pace of generations for the physical world, so too must our understanding and
change. Moreover, it is often challenging to make definitive decisions rules for the digital world. He writes, “The most critical next step in
about policy and technology when you are in the midst of an our pursuit of AI is to agree on an ethical and empathic framework for
industrial revolution. its design.”19

The countries that thrive in this climate are Through the experience of our customers, we’ve explored what AI
is accomplishing in the manufacturing supply chain today and the
those that enable the right framework of implication for the talent pipeline, and we also need to think about AI’s
broader societal and ethical implications. This chapter explores these
laws, policies, and principles to realize the implications and public policy considerations that can help create the

potential of these new technologies while right conditions for responsible innovation.

ETHICAL AI IN MANUFACTURING
also curbing their negative effects. In the
Microsoft has listened carefully to what a diverse group of stakeholders
era of the Fourth Industrial Revolution, AI in manufacturing have to say. From workers on the factory floor to

is no different. executives in the C-suite, and from labor unions and academics to
industry advocacy groups, what has come through all these conversations
But, as we have witnessed with previous technology paradigm shifts— is a sense of shared responsibility to create trusted, responsible, and
from the plow to the harvester, from the carriage to the automobile, from inclusive AI systems. We’ve heard that it’s time to work together to
the mainframe to cloud computing—nations have prospered when they reach a consensus about what principles and values should govern the
address these issues earlier, even if the solutions aren’t immediately clear. development and use of AI in manufacturing and that these values need
The countries that thrive in this climate are those that enable the right to reflect the experiences of organizations that are using AI every day.
framework of laws, policies, and principles to realize the potential of
Back in early 2018, in The Future Computed: Artificial Intelligence and
these new technologies while also curbing their negative effects. In the
its Role in Society, we wrote that designing trustworthy AI requires
era of the Fourth Industrial Revolution, AI is no different.
creating solutions based on ethical principles “deeply rooted in important
and timeless values.”

73 74
 Their responses raised a spectrum of issues from the risk of premature or
overzealous regulation to workplace safety, workforce development, data FOSTERING

privacy and security, consumer protection, and product liability laws. RESPONSIBLE
INNOVATION

Equally, much of the conversation revolved around the machine-to-

human relationship and ethical considerations about keeping “humans
in the loop.”

We were surprised and encouraged by the depth of our customers’

thinking about AI’s ethical dimensions, and their responses have
prompted us to think carefully about how the six principles for ethical AI
might apply to manufacturing.

In the book we proposed some principles we thought would assist Indeed, when we developed these principles our thinking was very much
in guiding the ethical development and deployment of AI. These six focused on our own role as a creator of AI technologies and how our AI
principles called for AI systems to be, reliable and safe, private and products and services could win the trust of customers and policymakers.
secure, fair, inclusive, transparent, and accountable. But as we conducted the research described in this book, it solidified
our assumption that these same principles should and do apply to any
At the time we published these principles, the discussion on AI ethics was
organization that uses AI, and, in particular, how these principles might
still in its relative infancy and there was not what we would characterize
apply to the manufacturing sector.
as a deep discourse on what responsible AI might mean for specific
sectors such as manufacturing. Back then, the most debated topic focused And while we believe the principles are broadly applicable, we also
on autonomous vehicles: How safe and reliable are the systems? How do recognize they will have different weight and relevance in different
they make decisions and on what basis? What data do they collect about segments of the manufacturing sector. For instance, the principle of
the driver? Who’s liable when things go wrong? privacy will have more relevance for the producers of smart elevators
than the producers of milk cartons.
Today, however, the discussion has somewhat matured. The debate on
ethics and AI is no longer limited to the end product of manufacturing Moreover, these principles are not about meeting some rigid checklist;
like self-driving cars—it now covers the entire digital value chain from they are about providing a flexible framework to promote the
design and engineering, planning, supply chain management, factory coexistence of business value and ethical values. As AI systems become
automation, and workforce training to IoT. more mainstream, we begin to understand that society has a shared
responsibility to create trusted AI systems and that collaboration is
During our exploration of AI in manufacturing, we asked leaders about
essential to reaching consensus about the principles and values that
their attitudes toward the concepts of responsible innovation and ethical
should govern AI’s development and use.
AI. Very often, unprompted, manufacturers said, “We need to
be extremely ethical.” Every one of them was deeply thoughtful about Let’s explore how each of these six principles might be applied to the
this issue. manufacturing sector.

75 76
 Safety and Reliability

Perhaps the most critical principle in manufacturing scenarios is safety FOSTERING

RESPONSIBLE
INNOVATION
and reliability. Every customer we spoke to put employee safety and
product reliability at the very top of their list—the complexity of AI has
fueled fears that it may cause harm in unforeseen circumstances. We
see this concern arising in areas ranging from autonomous robots on the
factory floor to autonomous vehicles on public roads.

As with any technology, however, trust will depend on whether AI-based

systems can be operated reliably, safely, and consistently over time and
not just under normal circumstances—they need to be safe, predictable,
and reliable in unexpected conditions or when they are under attack.
In manufacturing, concerns about safety and reliability take on a more
urgent tone, especially as factories become increasingly digital.

In a world where factory equipment is increasingly more complex,

automated, and interdependent, it’s not surprising that many of our
customers see the opportunity to harness AI to make their workplaces
safer and their processes more reliable. AI is increasingly important in
predictive maintenance for equipment, with sensors tracking operating
conditions and performance of factory tooling, predicting breakdowns
and malfunctions, and taking or recommending preemptive actions.

So, when developing AI systems for use in the manufacturing process,

particular care should be taken to ensure the safety of the people who
will be interacting with the system. In addition, AI systems should be
trained to function reliably in real-world situations and respond to
unforeseen circumstances. For example, an object recognition system that
inspects physical parts on an assembly line for manufacturing defects
should be trained to function in low-light environments, even if normally
it will operate with regular lighting. As another example, autonomous
robots that move around the factory floor could be designed to
recognize spills on the floor or respond to voice commands in the
event of an emergency.

77 78
Cybersecurity is also a key component of ensuring the reliability and The importance of safety and reliability goes as much for the factory
safety of AI systems. When AI systems are used in the physical world, floor as it does for the end products that consumers and citizens use FOSTERING

cybersecurity is critical to protecting people’s safety. For example, a and interact with. Industry groups and standards organizations can RESPONSIBLE
INNOVATION

cyberattack that incrementally slows the speed of a turbine on a cooling play an important role in helping share and promote best practices
system could result in overheating of the plant and a major safety and information sharing across the sector. It’s clear from our own
incident. Similarly, a facial recognition solution for monitoring employee conversations with customers that they are eager to share their insights
access to sensitive areas of a manufacturing facility needs to have robust and learnings to instill more confidence in the reliability and safety
measures in place to protect against malicious interference. of AI systems.

Beyond the factory floor, cybersecurity is critical for the safety of Privacy and Security
consumer products that have AI embedded. The stories of hacked baby-
Every stage of the manufacturing process involves humans in one way or
cams are certainly concerning, 20 but cyberattacks that compromise the
another. After all, at the end of nearly every manufactured product is a
software in digital pacemakers or automatic insulin pumps can mean
human user or consumer. But enabling the benefits of AI in applications
the difference between life and death. Regulators, such as the U.S. Food
that affect people requires access to large amounts of data about people.
and Drug Administration, are becoming more aware of this issue and are
strengthening the cybersecurity requirements for a range of consumer And just as humans are part of every stage of manufacturing, data is a
products, including medical devices. 21 needed at every stage of an AI solution: from developing and training
the AI algorithms to implementing and monitoring the AI system. Some
These issues of safety and reliability bring us back to the ongoing debate
examples from our customers and our own experience underscore the
about the relationship between humans and machines. AI systems should
ever-expanding role of data in intelligent manufacturing. But how to
be designed to be human-centric; namely, so people are always in the
protect privacy and secure the personal data that is used in AI systems is
loop and can take control when necessary. Indeed, some of our customers
a critical issue that is nowhere more important than in manufacturing.
not only seek to put humans in the loop but have developed language and
terminology about autonomous systems that underscores human primacy
over machines.

Ultimately, creating safe and reliable AI is a shared responsibility. It is

critically important for industry participants to share best practices for
design and development, such as effective testing, the proper structure of
trials, and reporting.

There is perhaps no better place to conduct such testing and trials than
in the factory setting. The expertise and discipline that manufacturing
brings to ensuring a safe workplace and reliable equipment could be
essential to learning how AI will apply in the real world.

79 80
AI systems used during the manufacturing process can directly or Fairness
indirectly capture personal data about people on the factory floor. AI
When we talk about AI and fairness, we mean that AI systems should FOSTERING

solutions designed to measure the efficiency and proper functioning RESPONSIBLE

INNOVATION
treat everyone in a fair and balanced manner and not affect similarly
of equipment, for instance, may yield information about individual
situated groups of people in different ways. Understanding why AI
employees’ efficiency and performance, while AI solutions that include
systems can behave unfairly and in what ways, as well as understanding
a microphone functionality (for example to monitor ambient noise or
who is most likely to be at risk of experiencing unfairness, is fundamental
to respond to voice commands) may inadvertently capture employees’
to operationalizing this principle. While there is no single definition
private conversations. Therefore, when using such systems, employers
of fairness that applies to all AI systems in all contexts, a variety of
should carefully consider the potential privacy impacts on employees and
processes and tools are emerging to help detect and mitigate unfairness
make sure that the use of these systems fully comply with all applicable
throughout the AI development and deployment lifecycle.
laws, including employment laws. Engaging with stakeholders, such as
employee representatives, can also be helpful to socialize new systems Fairness is particularly important where AI systems are used to make
and encourage uptake. decisions that impact people, something that could happen during the
manufacturing process in a variety of ways. For example, a company
We anticipate that manufacturers will be increasingly using AI and
could use an AI system to monitor workers’ alertness levels when
vast amounts of data to drive insights. These insights are not limited to
operating heavy machinery by reading their facial expressions, or to
running their manufacturing processes and their supply chains; they will
optimize efficiency by providing real-time feedback and insights based
inform them about their end products, the services bundled with these
on factory workers’ location and activities on the factory floor.
products, and how human users interact with them. Accordingly, when
making use of personal data, AI systems must comply with applicable
privacy laws, including both sector-specific laws and regional laws such
as the EU’s General Data Protection Regulation (GDPR), which has broad
extraterritorial reach.

Privacy laws generally require transparency about the collection, use, and
storage of data, and mandate that consumers have appropriate controls
so they can choose how their data is used. AI systems should also be
designed so that personal data is used in accordance with established
privacy practices and is protected from bad actors who might seek to
steal this personal data or inflict harm. In order to facilitate compliance
with privacy laws when sharing personal data to develop or deploy AI
systems, industry processes should be developed and implemented for
the following: tracking relevant metadata about personal data (such as
when it was collected and the terms governing its collection); accessing
and using shared personal data; and auditing access and use.

81 82
In each case the AI solution needs to have appropriate training data, Fairness becomes even more important when you consider the
model definitions, and fairness criteria to help ensure that it makes implications for manufacturing products that have embedded AI FOSTERING

appropriate, fair decisions and operates within appropriate parameters. solutions. Outside the production phase, fairness is an important RESPONSIBLE
INNOVATION

Training data that doesn’t reflect the real world—or systems that foundation of the product design and engineering stage because it is
inadvertently incorporate biases present in the real world because important to consider the ethical implications of products that can make
they are embedded in the training data or inherent in the decisions of decisions that might affect people in real-life scenarios. For example, the
the system’s designers—are common causes of AI bias. In the example manufacturer of smart doorbell cameras that recognize frequent visitors
above, a facial expression reading system trained on a dataset composed and flag suspicious activity would want to take particular care to ensure
primarily of images of men, or using model assumptions optimized for that the doorbell does not make unfair or discriminatory inferences such
male features, might have trouble properly interpreting the expressions of as those based on gender or race.
female employees.
Decisions made by teams at every stage of the AI development and
AI systems should also be designed to operate within appropriate deployment lifecycle can lead to unfairness. Therefore, much like security
parameters, so they don’t inadvertently make unfair decisions about and privacy, fairness cannot be treated as an afterthought or a “bolt-on.”
employees or put employees’ safety at risk. For example, an AI system Teams should implement processes to identify possible causes of unfair
that helps inform performance reviews or employee remuneration behavior—including societal biases—potential impacts on the people
should be designed and reviewed regularly to ensure that it doesn’t who will use or be affected by the system, and appropriate approaches to
inadvertently take into account factors outside the employees’ control mitigate these impacts.
(such as gender), doesn’t directly or indirectly produce outputs that are
Moreover, to help avoid bias and advance fairness, those designing AI
based on or influenced by irrelevant characteristics (such as age), and
systems should reflect the diversity of the workplace and the marketplace,
has enough flexibility and “breathing space” built in to allow for workers
and these teams should solicit input from diverse stakeholders during the
with different capabilities (such as poor vision but strong dexterity) to be
design process. In addition, if the recommendations or predictions of AI
considered fairly. In particular, supervisors or senior managers making
systems might be used to inform consequential decisions about people,
decisions based on AI systems should understand the limitations of the
it is critical that humans are “in the loop” on those decisions and that a
system and should not assume that these systems are more accurate and
designated person or people are accountable for these decisions. Those
precise—or more sensitive to context—than they actually are.
with accountability must be trained to understand the limitations of the
AI system, and to know how to intervene when necessary. In addition,
Fairness becomes even more important if AI systems are designed to evolve over time, it will be important to
when you consider the implications ensure that the fairness of the system is kept under continuous review,
taking into account the context in which the system operates.
for manufacturing products that have
embedded AI solutions.

83 84
Inclusiveness

Inclusive AI means that everyone should benefit from intelligent

technology, especially those with different physical capabilities. The
manufacturing workforce today is more socially and racially diverse
than many other sectors, but not so when it comes to age, gender, or
physical capabilities. For manufacturers, improving inclusiveness means
providing the workplace with a range of AI solutions that will allow for
more inclusive hiring and retention. This is important in many countries
where the demographics of an aging workforce means that those working
in the factories or on the shop floor are likely to be more reliant on
assistive technologies (e.g., technologies helping aging workers with
hearing and vision impairments or mobility limitations). In the United
States, the median average age of workers in the metalworking and
machine manufacturing industries is close to 49 years, with nearly 60%
of the entire workforce over the age of 45. 22 The situation is even more
acute in Japan—a manufacturing powerhouse—where in the last 20 years
the number of working-age adults has shrunk by 13% and in the next
20 years more than 30 percent of the population will be over 65 years of
age. 23 What this means is that if countries like the United State and Japan
are to maintain their competitiveness in manufacturing, companies must
accommodate an aging workforce with assistive AI technologies.

Thus, when we think of AI in manufacturing, we should not think

exclusively in terms of such efficiency-enhancing approaches as factory
automation, preventive maintenance, or intelligent supply chains. We
should also think about how AI can improve the overall well-being of
this diverse workforce.

Two areas to consider are healthcare and ongoing education. Because of

its diversity and the nature of the tasks it performs, the manufacturing
workforce has healthcare needs that are not the same as those of, say,
software developers, business executives, or lawyers. Embedded AI in
employer-provided health insurance and wellness programs can provide
unique new benefits tailored to the needs of factory employees.

85 86
Similarly, as discussed elsewhere in this book, ongoing reskilling and Transparency
upskilling is now looming as a critical issue for the manufacturing
Underlying the preceding values are two foundational principles that FOSTERING

workforce. While in one sense AI can be said to put some manufacturing RESPONSIBLE
INNOVATION
are essential for ensuring the effectiveness of the rest: transparency
jobs at risk, it can also be part of the solution by providing new teaching
and accountability. Let’s begin with transparency.
tools to enhance workers’ skills and help them transition to new roles.
When AI systems are used to help inform decisions that have tremendous
Manufacturers should consider how their impact on people’s lives, it is critical that people understand how those
decisions were made. Transparency improves the quality of products
products, including those embedded and engenders trust among manufacturers, customers, partners, and

with AI systems, could be used by a regulators, ultimately leading to wider adoption. In part, transparency
means that those who build and use AI systems should be forthcoming
diverse range of people, of different ages, about when, why, and how they choose to build and deploy their systems,
as well as their systems’ limitations. Transparency also means that people
genders, or physical capabilities. should be able to understand and monitor the technical behavior of AI
systems—what we refer to as “intelligibility.”
Inclusiveness is also important beyond the factory floor, in the wider
society where the manufactured products are used. Manufacturers The need for intelligibility can arise in an almost limitless range of
should consider how their products, including those embedded with scenarios involving any number of human actors. And the methods for
AI systems, could be used by a diverse range of people, of different achieving intelligibility can be just as varied. In some cases, information
ages, genders, or physical capabilities. Making sure that technology is about the development pipeline, the relationships between different
inclusive and accessible to everyone is important to ensure fair access. system components, or performance metrics may be enough to help
In Microsoft’s 2018 book, “The Ability Hacks”, Microsoft developer Guy stakeholders achieve their goals; in other cases, such as those involving
Barker implores fellow developers to: system developers, more detailed information about system components
may be required. For example, if an AI system corrects a machine that
“Consider how each feature might help someone interact with your app.
may be shifting and drifting, manufacturers should be able to look
And as you build up your processes to design, build and test products
behind the data and algorithm to determine in detail why and how
that every person can efficiently use, always keep in mind the rationale
the correction was made. Similarly, for the manufacturer deploying the
for doing so. It’s a question of fundamental fairness. No person should be
system, it’s not enough simply to receive a notification about a predicted
prevented from being connected with others, or from being employed,
failure—the manufacturer needs to understand how the AI reached that
simply because apps block them from those core aspects of life. And
conclusion. Ultimately, it is the manufacturer, not the machine, that is
what’s more, with the tools at your disposal, it’s practical for you as an
accountable, and for the manufacturer to be accountable, they need to
individual who cares about quality for all to build an accessible app.”24
understand how the AI system works.

87 88
In other cases, stakeholders such as workers and consumers may need particularly important when AI systems are used to make consequential
more basic information about the operation of an AI system in order to decisions, such as selecting suppliers, carrying out performance reviews, FOSTERING

interact with it in an accessible and useful way. or operating key functionalities in consumer products. It is important RESPONSIBLE
INNOVATION

that people know that AI systems are being used and have information
This builds trust in AI systems and enables people to identify potential
about the purposes and limitations of these systems. Among other things,
performance issues, safety and privacy concerns, biases, exclusionary
transparency ensures that ongoing feedback can be obtained from people
practices, or unintended outcomes. For example, an agricultural AI
who interact with AI systems, so that people can get in touch if they see
solution may be used to help farmers select strains of seeds to plant on
errors or have concerns. This is important to ensure AI systems continue
the basis of predicted weather patterns and soil quality. The farmers who
to function properly and safely as well as foster trust in the AI system by
make use of the AI solution do not necessarily need to understand the
the people who use it.
technical models underlying the algorithm, but they should understand
the factors that the AI system took into account and the limitations of Accountability
the methodology adopted so they can make informed decisions. For
The final, and perhaps most important, principle is accountability. In
instance, the AI solution may not have accounted for other factors that
short, the people who design and deploy AI systems must be accountable
may be relevant to the farmer, such as whether the seeds were sustainably
for how their systems operate.
sourced, the impact of the seeds on the quality of next year’s soil, or the
health of local pollinator populations. Manufacturers should draw upon industry standards to develop
accountability norms for their own organizations. These norms can help
In short, the people who design and ensure that AI systems are not the final authority on any decision that
impacts people’s lives, and that humans maintain meaningful control
deploy AI systems must be accountable over otherwise highly autonomous AI systems, especially when AI

for how their systems operate. systems make consequential decisions. In order to ensure that people
remain ultimately accountable for AI systems and their operation, those
A number of promising approaches to achieving intelligibility have who use AI systems on a day-to-day basis should be trained to review
begun to emerge. Some facilitate understanding of key characteristics of results, identify errors, and understand the limitations of the AI system.
the datasets used to train and test models. Other approaches focus on They also should have the authority and training to take any remediation
explaining why individual outputs were produced or why predictions steps necessary.
were made. Others offer simplified, but human-understandable,
Organizations should also consider establishing a dedicated internal
explanations for the overall behavior of a trained model or an entire AI
review body. This body can provide oversight and guidance to enterprise
system. We encourage further research into these and other approaches
leaders on which practices should be adopted to address the concerns
to help expand the range of tools available to address the diverse
discussed above and on particularly important questions regarding
requirements of all stakeholders.
the development and deployment of AI systems. They can also help
We also believe that those who use AI systems should be transparent with defining best practices for documenting and testing AI systems
about when, why, and how they choose to deploy these systems. This is during development, as well as providing guidance when an AI system

89 90
 will be used in sensitive cases, such as decisions that may deny people
consequential services like healthcare or employment, create risk of FOSTERING

physical or emotional harm, or infringe on human rights.At Microsoft, RESPONSIBLE

INNOVATION

we take accountability very seriously. Our Senior Leadership Team

established the AI and Ethics in Engineering and Research (AETHER)
Committee to serve in an advisory role on rising questions, challenges,
and opportunities with the development anddeployment of AI
and related technologies. In addition to its work in support of the
principles above, the committee is charged with providing insights and
considerations on a broad range of policies, processes, and best practices.

Ultimately, these principles must be

reinforced by public policy, laws, and
standards to ensure that they are evenly
applied and to build both confidence
in the technology and trust in those
who create it.

NEW RULES FOR NEW TECHNOLOGY

While principles and best practices can be important safeguards in

the absence of laws or regulation, it is clear that more formal policy
development is also needed to help create the right environment for
responsible innovation in manufacturing to flourish.

During our exploration of AI in manufacturing, we asked leaders about

their own attitudes on policy and regulatory priorities. Every customer
we spoke to was already thinking about the role of policy in shaping the
use of AI in manufacturing.

91 92
Not surprisingly, their attitudes ran the spectrum of issues that reflects Customers want a consistent regulatory
the diversity of the manufacturing organizations they represented. What
framework on how to share data with
FOSTERING

emerged, however, were a few consistent themes: RESPONSIBLE

INNOVATION

• The importance of a regulatory regime that fosters confidence in AI, confidence, both internally and across the
both by those who invest in AI technologies and those who interact
with it. world, particularly given the increasingly
• Ensure that there is equitable access to AI and, most importantly, interconnected global supply chain.
access to the critical data that will fuel these AI-led innovations.
Organizations that invest heavily in AI technologies also want to ensure
• Capacity-building for organizations to leverage AI and the workers that they can retain effective control and ownership over the proprietary
who will be working with AI. data that gives their AI systems a competitive edge in the market. As one
customer noted: “We [have] legal limitations when it comes to external
• Collaborative approaches to the creation of the rules, standards,
use of data … [but] there is very little to no limitations inside the plant …
and laws that balance innovation with responsibility.
it is more of an ethical question in terms of how much we can use to train
• The importance of a regulatory regime that fosters confidence in AI, machines. There needs to be clear regulation that ensures data will be
both by those who invest in AI technologies and those who interact kept private. Ownership of the data is very important. A customer’s data
with it. stays their data.”

• Encouraging the use of AI to create a more environmentally In addition, manufacturers also are looking carefully at the future of
sustainable supply chain. consumer protection and product liability laws. In a world of data and
AI-driven decision-making, who is liable when things go wrong: the
Below we will explore each of these in more detail:
operator or the creator of the AI? How can manufacturers, developers,
Fostering Confidence and operators work together to apportion liability to ensure that
consumers have adequate protection when something goes wrong, while
The direction of policy developments on issues of data privacy and
also ensuring that incentives to innovate are reasonably preserved?
security, intellectual property rights, consumer protection, and liability
are universal concerns raised by customers with whom we met. Although frameworks that assess the reasonableness of the conduct
They wanted a consistent global regulatory framework that fosters of relevant actors are well suited, some frameworks, such as strict
confidence in AI for those who invest in AI technologies and those who manufacturer’s liability, may not always lead to the most equitable
interact with it. In the area of privacy and security, customers want a outcome when AI systems are involved. Since multiple parties can be
consistent regulatory framework on how to share data with confidence, involved in the development and deployment of AI systems, it will be
both internally and across the world, particularly given the increasingly important to ensure that liability systems reasonably apportion liability
interconnected global supply chain. for any defects or other harms. It will also be important for industry
actors to work closely with insurers to make sure that the risks posed by
AI systems are properly insured.

93 94
Finally, it should be remembered that AI systems do not operate in
a regulatory vacuum. AI systems are subject to a number of existing FOSTERING

laws, including privacy laws, product liability laws, consumer laws, RESPONSIBLE
INNOVATION

anti-discrimination legislation, and sector-specific laws. Existing data-

ownership models, such as trade secret protection, may also apply
to some aspects of AI systems. Governments should assess existing
regulatory frameworks that apply to AI systems and consider both where
new measures are needed and where existing legal requirements, or a
lack of clarity in what the law requires, pose unnecessary impediments to
AI innovation.

Importantly, any new regulatory frameworks on AI should be consistent

with existing laws. Industry should work closely with governments to
provide practical guidance on the direction for future policies on AI Perhaps more striking are the comments by experts who worry that
regulation, and developing industry standards can be a good starting the concentration of data in the hands of a few large companies could
point for establishing consensus in the market. threaten AI innovation and pose a challenge to equal opportunity more
broadly. It is vital that we get the fundamental tools and resources of data
Equitable Access
and AI into the hands of as many people as possible.
Democratizing access to AI is a key priority for stakeholders. They
One way to address this information asymmetry is to have policies that
remarked that public policy is necessary to create the right incentives to
expand “data commons” efforts across the manufacturing supply chain.
help all organizations, regardless of size, benefit from AI’s potential.
This would provide organizations large and small with access to rich and
They agreed that equitable access to AI, across the entire supply chain, diverse raw datasets that can be used to help refine and train their AI
would have a multiplier effect with greater efficiencies, deeper insights, solutions. Such efforts bring together collaborators to address issues that
and speed to innovation being shared by all. none of the entities alone can tackle.

Consistent with this concept of equitable access was the key role of Governments can also help accelerate AI advances by promoting
data; namely, that if the necessary training data is only available to common approaches to making public-sector data broadly available
those organizations that control the largest data estates, then this could for machine learning. A large amount of data resides in public
disadvantage smaller manufacturers. datasets managed by governments that could be exceedingly useful
to manufacturers, such as climate data, land survey data, or datasets
As one Microsoft executive explained, “Data are as fundamental to the
created from publicly funded research. Governments can also invest in
AI revolution as fossil fuels were to the industrial revolution. Unlike fossil
and promote methods and processes for linking and combining related
fuels, however, the challenge with data is not scarcity but access.”
datasets from public and private organizations (on a voluntary basis)
while preserving confidentiality, privacy, security, and ownership rights
as circumstances require.

95 96
Any efforts to create such data commons would need the right data Through the creation of an open technology framework and cross-
protection and privacy rules, licensing and data sharing agreements, industry community, the OMP is expected to support the development FOSTERING

protection for intellectual property and standards, and finally, of smart factory solutions that will be shared by OMP participants RESPONSIBLE
INNOVATION

standardized format or schemas to ensure broad usability. across the automotive and broader manufacturing sectors. The goal is
to significantly accelerate future industrial IoT developments, shorten
A recent example of sharing data across the technology industry is the
time to value and drive production efficiencies while addressing common
Open Data Initiative (ODI) between Microsoft, SAP, and Adobe. 25 This
industrial challenges.
partnership is designed to improve customer experience management by
empowering companies to derive more value from their data and deliver Built on the Microsoft Azure industrial IoT cloud platform, the OMP is
world-class customer experiences in real time. The ODI is a intended to provide community members with a reference architecture
common approach and set of resources for customers based on with open source components based on open industrial standards and an
three guiding principles: open data model. In addition to facilitating collaboration, this platform
approach is designed to unlock and standardize data models that enable
• Every organization owns and maintains complete, direct control of
analytics and machine learning scenarios—data that has traditionally
all their data.
been managed in proprietary systems. Utilizing industrial use cases
• Customers can enable AI-driven business processes to derive and sample code, community members and other partners will have the
insights and intelligence from unified behavioral and capability to develop their own services and solutions while maintaining
operational data. control over their data.

• A broad partner ecosystem should be able to easily leverage an With currently over 3,000 machines, robots and autonomous transport
open and extensible data model to extend the solution. systems connected with the BMW Group IoT platform, which is built
on Microsoft Azure’s cloud, IoT and AI capabilities, the BMW Group
Based on these principles, every organization owns and maintains
plans to contribute relevant initial use cases to the OMP community.
complete, direct control of all their data. The core focus of the ODI is
One example is the company’s use of their IoT platform for the second
to eliminate data silos and enable a single view of the customer, helping
generation of its autonomous transport systems in the BMW Group
companies to better govern their data and support privacy and security
plant in Regensburg, Germany, one of 30 BMW Group production and
initiatives. With the ability to better connect data across an organization,
assembly sites worldwide. This use case has enabled the BMW Group
companies can more easily use AI and advanced analytics for real-time
to greatly simplify its logistics processes via central coordination of the
insights, “hydrate” business applications with critical data to make
transport system, creating greater logistics efficiency. In the future, this
them more effective, and deliver a new category of AI-powered services
and other use cases—such as digital feedback loops, digital supply chain
for customers.
management and predictive maintenance—will be made available and, in
Another example is the Open Manufacturing Platform (OMP) which fact, developed further within the OMP community.
Microsoft and the BMW Group annouced at Hannover Messe in April
The ODI and OMP are just two examples of companies working together
2019. The OMP is a new community initiative to enable faster, more cost-
to derive better insights from diverse supply chain information. This can
effective innovation in the manufacturing sector.
provide a model for other industries and companies to follow.

97 98
 “Business leaders are aware of the massive reskilling efforts required to
build an AI-ready workforce. However, 22 percent of business leaders FOSTERING

felt that workers have no interest in reskilling, but only 8 percent of RESPONSIBLE
INNOVATION

workers feel the same. In addition, 48 percent of business leaders feel

that workers do not have enough time to reskill, but only 34 percent of
workers feel the same way.”

What this means is that business leaders must prioritize reskilling and
upskilling, dedicating employees’ time for this to address the skills
shortage. Even though it may result in short-term productivity impact,
building an AI-ready workforce will result in greater gains in the future.
For policy makers, a mobile and more dynamic workforce will increase
pressure on the already constrained social safety net which will need to
Capacity Building anticipate changes to unemployment insurance, reemployment programs
– including job training and trade adjustment programs – and paid time
Many business leaders and workers in the manufacturing sector believe
for training.
that AI will have a positive impact on their jobs and organizations. This
is a sentiment backed up by a recent study by IDC of business leaders With the customers we spoke to directly, some wanted to see support for
and workers from across 15 countries in Asia Pacific. 26 The manufacturers “earn and learn” strategies. “We need a societal approach,” said one.
who were interviewed spoke passionately about the urgency of expanding
educational opportunities and skills development related to technology. “Readiness implies continued learning.
The advent of AI has only increased this sense of urgency, and not only
for technical skills. The business leaders interviewed identified the top We won’t put people out of work,” said
three skills where demand will outstrip supply in the next three years: 1)
another, “but we want to make sure they
communication and negotiation skills; 2) entrepreneurship and initiative-
taking; and 3) adaptability and continuous learning. are trained to have the right job.”
At the same time, business leaders believe that demand for basic data Another manufacturer was blunter: “We’re going to displace people.
processing, literacy and numeracy, and general equipment operations and People need to know that. We need some planning around that. What
mechanical skills will decrease in the next three years. Those surveys skills are needed? What is the next step?”
claimed those skills are broadly available today, and already now the
One executive recalled that the sector leaped onto the “Lean
supply is higher than the demand in Asia Pacific.
Manufacturing” bandwagon. Suddenly they realized people can lose jobs
As Scott Hunter, Microsoft’s Regional Business Lead for Manufacturing, by focusing on waste reduction. “But they figured it out. I can’t change
commented, there is a disconnect between employers’ perception of their people’s mindset [about AI] if they fear it.”
workers’ willingness to reskill and the actual willingness of workers.

99 100
A Collaborative Approach There was also a strong view that governmental bodies and industry
must come together to agree on sensible standards. There is also the
There was a consensus that, given the relative immaturity of AI, FOSTERING

need to think in strategic terms at both a national and an industry RESPONSIBLE

INNOVATION
governments and policymakers need to avoid regulatory overreach,
level. “The standards are not fully developed,” one manufacturer told
which so often can foreclose innovation opportunities. Manufacturers
us. Having a rich set of AI standards will make it easier to scale training
encouraged a more light-touch regulatory regime in the short term while
for workers and make it easier to deploy technology.
governments continue to create a longer-term policy framework. As
one manufacturer said, “AI is at a low level of intelligence today. We are This was confirmed by a European company, which said that government
expecting it to do things it can’t. Its level of intelligence is like a mouse. needs to either embrace AI or be left behind. “Places with the highest
AI is only as intelligent as we make it.” concentration of robots have the lowest unemployment. Policy, academia,
and industry need to come together.”
Another customer observed, “If government tries to control, we will find
ourselves falling behind and miss the boat. In Europe, regulation is not One European policymaker stressed the need for harmonizing digital
empowering digitalization. Don’t regulate for control but for benefits.” laws worldwide: “We can’t adopt rules in our own corner and not talk
with one another.” Industry-led organizations such as the Partnership on
One executive noted that government does not yet understand AI
AI, 27 which brings together industry, nonprofit organizations, and NGOs,
in a way that would make their ideas constructive and helpful. “All
can serve as forums for the process of devising best practices and
governments are trying to catch up with the development of technologies.
entering into a dialogue with governments on the direction of future
Legislation is always slower than the tech itself.”
policy and regulation.
Beyond this light-touch regulatory approach, customers commented
One place where these various policy threads can be woven together is
on the need to look at policy development from a whole-of-industry
through national AI strategies. These blueprints serve as a cross-sector
perspective and avoid making policy based on narrow considerations.
and multi-stakeholder effort to align the various interests and needs
Given the pervasive nature of AI across all industries, stakeholders are
of the public and private sectors as well as serve the broader interests
anxious that policymakers understand this intersectionality earlier and
of the community. To date, more than 20 countries have announced
think about industry policy when they think about digital policy. “Do
comprehensive national AI strategies, outlining how AI can boost
something that makes sense,” one source cautioned. “Policymakers
national competitiveness while addressing societal concerns. 28
need to get to know the processes manufacturers use. Don’t create
impediments that would get in the way of connectivity of data. We are Many of these strategies put manufacturing and Industry 4.0 as a key
trying to improve process and make customers better, so don’t get in the driver and have set up ethics panels and high-level expert groups to
way of that.” address the economic, ethical, policy, and legal implications of AI. The
voice of business leaders, labor representatives, and technology providers
“One big fear,” said a senior executive at an American company, “is that
will be critical to realizing the vision behind these plans.
we will get into a regulatory situation about data sharing and exchange.
Bricks and mortar are no longer the asset of the company—data is.
Government wants to protect, but with protection comes control.”

101 102
 There is a globally recognized imperative to become more sustainable
and address the world’s most pressing environmental challenges. FOSTERING

Scientists say we have less than two decades to course correct the RESPONSIBLE
INNOVATION

impact humans are having on the planet, 32 and United Nations member
states issued an urgent call for action and a global partnership to tackle
this, including promoting responsible production. 33

The challenge is clear—manufacturers must break the pattern of past

industrial revolutions, and deliver not just economic growth, but
sustainable growth.

Sustainability also has significant reputational value. A report by

McKinsey & Company 34 notes that environmental sustainability is
Environmental Sustainability now a strategic and integral part of many businesses. Indeed, a Nielsen
Manufacturing and production activities account for a significant study of 30,000 consumers across the world indicates that 73 percent of
impact on the environment. Global manufacturing consumes some 54 Millennials are willing to pay extra for sustainable goods. 35
percent of the world’s total energy and creates around one-fifth of global Therefore, sustainable manufacturing is no longer optional—it’s a
greenhouse gas emissions. 29 Further, the global water crisis is exacerbated business imperative. Manufacturers need to leverage the best
by the almost insatiable need for water in the manufacturing of most technologies possible to harness vast amount of data and
goods, with millions of gallons of water being used each and every day make breakthrough advances in reducing environmental impact
to produce consumer goods. Just a few examples underscore this point: and managing scarce resources.
It takes between 13,737 to 21,926 gallons of water to make a car; 3,626
gallons for a pair of leather shoes; 3,190 gallons for a smartphone; 2,108 Manufacturers have already started leveraging AI for sustainable
gallons for a pair of jeans; and 659 gallons for a T-shirt. 30 production, ensuring it is economically and environmentally sustainable.
For example, companies such as Ecolab, Ørsted, and The Yield have
This growing water and energy consumption in manufacturing is set developed AI-enabled solutions to improve water conservation,
against an already challenging environmental situation: communities that renewable energy management, and agricultural production. A small
are ravaged by floods and wildfires; farmers who are losing their harvests startup, SilviaTerra, is using AI to improve forest management and
to pests and extreme weather; warming oceans, deforested wilderness, fuel the carbon offset market with precise data about the carbon
arid soil, and contaminated watersheds. As these crises unfold, they are sequestration potential of every acre of forestland in the United States.
devastating for local communities who go hungry, thirsty, or lose their
homes and livelihoods. Environmental challenges also impact the broader These companies’ experiences demonstrate that AI can enable
ecosystem, rippling across national boundaries and disrupting the global innovations in sustainable manufacturing and go hand in hand with
economy, costing billions of dollars in business and property damage, not profit-making and improved competitiveness.
to mention loss of human life. 31

103 104
These are not isolated examples. New research from PwC shows that S
to improve the efficiency of their production processes and products,
AI can be applied to a wide range of economic sectors and industries enabling them to contribute to sustainable development and green
to improve environmental outcomes and mitigate climate change—all growth. 38 The toolkit includes an internationally applicable common set of
while driving economic growth. 36 In fact, in their research, PwC estimates indicators to help businesses measure their environmental performance
that using AI for environmental applications could contribute up to at the level of a plant or facility. 39
US $5.2 trillion to the global economy in 2030, a 4.4 percent increase
There is enormous potential for AI to be an important tool in the effort to
relative to business as usual, and create an estimated 38.2 million new
decouple economic growth from rising carbon emissions. There is a clear
skilled, green jobs across the global economy. The report highlights how
path for manufacturers toward a prosperous and more sustainable future
AI is increasingly part of the suite of options in heavy industry, which
with advanced technologies.
alongside heavy transport, has often been called out as “hard to abate.”
This is the future of sustainable manufacturing and environmental
Additional examples include:
sustainability for our planet, and AI has a substantial role to play.
• Cement: AI, often combined with advanced sensors, is being Whether it’s deployed to monitor and improve environmental outcomes,
used for predictive asset management to maximize the efficiency, enhance resource management, or create a greener economy, taking a
operation, and management of production assets. tech-first, AI-enabled approach is the path forward.

• Steel: AI systems are increasingly being tried and tested in process Where to from Here
and operational controls to supplement traditional controls,
As with the great advances of the past on which it builds—from
increase efficiency, and optimize system operations.
the steam engine, to the combustion engine, from electricity to the
• Chemicals: AI, often combined with traditional controls and new microprocessor—AI will bring about vast changes to the manufacturing
sensors, is being deployed for predictive maintenance in chemicals sector, some of which are already being realized by organizations today.
manufacturing, maximizing efficiency and minimizing resource use
But just as with these previous significant technological advances,
in chemicals processes.
we’ll need to be thoughtful about how we address the economic,
• Shipping/supply chain: AI can be used in the maritime sector employment, societal, and environmental issues that these changes
a range of ways, including for predictive vessel management and bring about. Most importantly, we all need to work together—
maintenance, real-time voyage optimization, and fuel monitoring industry, government, academia, educators, civil society, and labor
and management. 37 representatives—to ensure that AI is developed and deployed in a
responsible and trustworthy manner.
And with the democratization of AI, small and medium-sized business
now have access to these same innovative technologies and are starting to Each of us has a responsibility to participate and an important role to
embrace these great opportunities. play. For our part, one of those roles is helping organizations of all sizes
realize their tech intensity vision. The next chapter is all about helping
They are learning from the earlier pioneers and are leveraging resources
customers on their AI journey and their pathway to innovation.
such as the “OECD Sustainable Manufacturing Toolkit”, which aims
to provide a practical starting point for businesses around the world

105 106
AI AND THE CONNECTED SUPPLY CHAIN:
THE MICROSOFT EXPERIENCE

Sitting in her office on the Microsoft campus, Pat Flynn-Cherenzia

glances at her computer screens and quickly assesses the company’s
supply chain all over the world. At the moment, shipments of
Microsoft supplies and inventory look mostly green along the
U.S. coasts. Normally at this time of year she can see problems in lose visibility. But with Microsoft’s cloud and cognitive services
Florida. It’s red in Europe due to ice, and outbound traffic in Asia is in place, Pat was able to use independent data to triangulate and
yellow as businesses rush to get product out of ports before real and predict where shipments were within 20 minutes after the attack.
threatened tariffs hit.
“We were able to tell them where their shipments were so they could
Pat uses AI systems to help predict when certain events will occur. be retrieved.”
She’s led logistics at Microsoft for years, but earlier in her career
she worked for U.S. Customs, the ultimate supply chain. Moving to the cloud has allowed Microsoft to leverage machine
learning and predictive analytics, which lets Pat anticipate
“So much is coming and going, you couldn’t inspect everything so disruptions. At the time of the Xbox launch, South Korea’s Hanjin
you had to combine datasets and create algorithms that could Shipping was offering attractive rates, but the firm abruptly filed
tell you, ‘this is a viable candidate for inspection or this is a for bankruptcy and it took a week to find Microsoft’s inventory.
trusted source.’” “Even if the provider doesn’t tell me, I know by SKU where my
A native of Philadelphia, Pat worked for U.S. Customs there, products are on boats. Satellites are tracking my goods. Based on
where her goal was to prevent intrusions at that city’s port. She trends, I can predict when the product will arrive at port. If I see a
was on the lookout for first-time importers from places where the weather pattern building in the Caribbean, and I see our product is
threat was elevated and the shipper had a “funky product,” like headed in that direction, I can hold it back.”
say a rickshaw from Thailand. Turns out that shipment was full of During Hurricane Michael in October 2018, Pat held back
narcotics, and she caught it. Fast forward to today, and Pat is a key shipments headed for the Florida panhandle and instead diverted
player in Microsoft’s own digital transformation in manufacturing. to Memphis just in time. Microsoft gives the same visibility to its
She oversaw multiple “control towers” for Nokia’s factories and customers, so in this case the retailer Best Buy could know with
distribution centers, speeding up the onboarding of new suppliers confidence when it would see its product.
like DHL by moving to a single platform in the cloud.
Pat has since moved onto another role within Microsoft to share the
Later, when a malware attack hit the shipping giant Maersk, it depth and breadth of her experiences with small, medium and large
decimated their systems, causing the shipper and its customers to manufacturing customers worldwide.

107 108
CHAPTER 4

THE WAY
FORWARD

109 110
THE WAY FORWARD To address these challenges, Microsoft has worked with customers in
all sectors of the economy to define an operational model that helps
organizations assess their own attributes and maturity. Microsoft’s AI THE WAY
FORWARD

Maturity Model helps organizations conduct this assessment and guides

adoption of the right kind of AI at the right place and time, smoothing
AI MATURITY AND PROGRESSING YOUR JOURNEY the path to safe deployment of progressively more advanced capabilities.

The AI Maturity Model includes four stages of development:

As we have seen, AI is a critical component in helping manufacturing
organizations transform their operations, better serve customers, and
offer new opportunities to their workers. To realize this potential,
each organization needs to adopt a strategy, culture, and set of core
capabilities around AI that match the organization’s maturity. The
following section looks at how organizations can progress their AI
journey in a way that corresponds to their maturity level.

AI-based systems must be

continuously trained, monitored,
and evaluated for performance if
organizations are to realize their full
benefits while guarding against bias,
privacy violations, and safety concerns
Deploying an AI-based system is very different than acquiring traditional Foundational
packaged software or developing a custom-coded non-AI solution. AI-
Organizations at the foundational stage are seeking to understand
based systems must be continuously trained, monitored, and evaluated
the varieties and applications of AI and how others in their industry
for performance if organizations are to realize their full benefits while
are using it.
guarding against bias, privacy violations, and safety concerns. Neglecting
this maturity assessment can seriously impede an AI project, potentially They strive to make more data-driven decisions and currently tend to
causing a rejection of the technology by employees who perceive it as too rely on the instincts of experienced leaders to make decisions.
difficult to use or untrustworthy.

111 112
Foundational organizations need to invest in projects that focus on fast, Aspirational organizations can adopt configurable AI, hosted by
iterative experimentation. Doing this successfully requires organizations technology companies. This abstracts away the operational complexity
to build a culture that embraces experimentation and empowers of maintaining the core AI while allowing organizations to infuse AI THE WAY
FORWARD

colleagues to make data-driven decisions. Organizations at this level into digital experiences. At the same time, experimentation with more
of maturity should look to adopt AI technologies built on established advanced AI technologies such as custom AI is encouraged to these
platforms, helping them to grow into digital businesses. organizations learn about how to operate and coordinate more
complex systems.
Approaching
Mature
Organizations at this level of maturity are implementing cultural changes
to empower employees and make data-driven decisions. They are Mature organizations have shifted their culture to embrace rapid,
focused on building a data culture and using AI to build new processes iterative experimentation and a data-driven approach.
and streamline operations. Having digitized assets and deployed AI to
Mature organizations develop AI talent and understand how to apply this
automate certain processes, these organizations are ready to explore
resource to multiple AI initiatives simultaneously.
owning custom AI solutions. Approaching organizations are poised to
embrace rapid experimentation and will invest more in understanding Mature organizations ask not just what can we do with AI but what
how to implement and improve AI over time. should we do with AI. The organizations also infuse ethical perspectives
into their experience creation process.
Investments should continue in accountability protocols for AI
governance, monitoring, orchestrating, and improving AI over time
and infusing ethical viewpoints in AI-based systems. Considering
Mature organizations ask not just what
these issues will help organizations gain experience when using AI to can we do with AI but what should we
digitally transform.
do with AI. The organizations also infuse
Aspirational

Aspirational organizations understand that AI will be instrumental in

ethical perspectives into their experience
helping them compete and transform. These organizations know creation process.
that others are using AI and understand the competitive disruption
this may entail. Organizations at this level of maturity should continue to evaluate
tool chains for configurable and custom AI while being vigilant about
Organizations at this maturity level are focused on shifting culture to monitoring, retraining, and updating AI-based systems. Maintaining
empower employees, increasing collaboration, and generating ideas for AI talent, prioritizing new strategic initiatives, and continued agile
optimization, new offerings, and business models. These organizations experimentation are areas of focus for mature organizations.
are becoming increasingly comfortable with taking risks and are striving
to transition away from fixed projects to more iterative projects.

113 114
 A Path to Innovation

Digital transformation is a journey. It provides a path for organizations THE WAY

FORWARD
to innovate and establish new and better ways of doing business that
benefit all stakeholders.

Considering and assessing an organization’s AI maturity provides a clear

path to guide AI technology adoption efforts. Organizations closer to the
Foundational or Approaching stages should look to adopt configuration-
based AI first, where the concepts of operationalization are developed by
partners like Microsoft. When more specific AI capabilities are required,
organizations should assess themselves across strategic, cultural, and
capability boundaries to determine if they are ready to own and operate
a custom AI solution.

Microsoft works with customers to prepare, host, and implement

configurable and custom AI to help organizations digitally transform.
Through a combination of AI services, AI platform capabilities, and
smart digital experiences, we work to make the infusion of AI into digital
experiences a rapid, iterative process that speeds transformation.

Enterprise leaders should recognize that we are all transitioning from

an era when every company is a digital company to one where every
company must now become an AI company. And manufacturers now
need to be thinking as digital and AI companies. This will be a cultural
shift, as well as a technical and business one.

115 116
CONCLUSION Historical data might show that a certain drop in electricity generated by
a solar panel may indicate that a panel needs to be cleaned or a battery
As this book went to press, more than 200,000 manufacturers and 6,000 checked within 12 hours or it could fail. The analytics allow remote CONCLUSION

exhibiting companies converged on the German city of Hannover for the monitors to help proactively ward off those types of problems.
annual tradeshow, Hannover Messe.
Another of our customers at Hannover Messe was Repsol, one of the
On display at the tradeshow were the very latest products and trends in world’s largest oil and gas exploration and production companies. They
industrial technology. Mile after crowded mile, in cavernous halls the spoke to us about how AI is helping to fuel the world’s energy demands
size of airplane hangars, manufacturers from every corner of the globe while also protecting the environment. Data and computing make it
were treated to stunning displays of ever-evolving automation—robots, possible to see beneath layers of soil, sequence DNA, and therefore
sensors, IoT breakthroughs, AI and machine learning, drones, and every make drilling more precise; minimizing not just the number of wells
manner of drill, assembly line, and factory floor machine. needing to be drilled, but maximizing the time to drill them. All of this
helps reduce the time from acquisition of the site to when the energy is
It’s clear that Industry 4.0 has arrived. commercialized.

Manufacturers are now digital companies. Every manufacturer with whom we engaged reported that AI is boosting
efficiency, productivity, safety, and health. Those policymakers we
While the machinery on show was impressive, it’s what you can’t see spoke to also reported that a stronger economy from a more efficient,
that takes center stage—data. It’s clear that Industry 4.0 has arrived. sustainable, and competitive manufacturing sector will lead to
Manufacturers are now digital companies. Not surprisingly, the effect improvements for their constituents and for the economy overall. The
was inspiring and just a little overwhelming. Not unlike the premise of economic impact was not the only issue on which manufacturers and
this book. policymakers agreed; they also were in agreement that an AI-led future
When we set out to write this book, we had one simple objective: listen requires a commitment to bringing along the current workforce and
to manufacturing customers and public policy experts to learn about cultivating a long-term skills pipeline for future generations.
their AI journey, report their stories, and offer our own insights and To underscore the implications for the next generation of talent, one
recommendations on the implications and possibilities that lie ahead. executive told a story about a recent tour of colleges for prospective
At Hannover Messe, one of our customer and partners, Schneider students where a career counselor asked the applicants what year they
Electric, was on hand to demonstrate how its platform is leveraging would graduate high school and then what year they would graduate
Microsoft AI to help a range of customers stay ahead of maintenance college. Students and parents alike were confident, and the answers
problems in applications as varied as coffee roasters in the developed came fast and furious. But when asked what year they would retire, the
world to schools and clinics in the developing world. audience fell silent. The counselor then estimated that, whenever their
retirement might be, the students would have worked in seven different
One example was their use of Microsoft’s Cortana Intelligence Suite fields, two of which had not yet been invented. In other words, there will
in Nigeria, where Schneider can identify trends in its solar panels so be disruption, and we must all prepare.
technicians can address issues before they lead to outages.

117 118
Fortunately, that is happening. According to LinkedIn Chief Economist
Guy Berger, the presence of AI-related certifications in American
manufacturing is higher than in the typical industry and rising, with Guy CONCLUSION

reporting that 0.86 percent of certifications in U.S. manufacturing are

AI related as of 2018. This is more than double the share seen two years
ago. While still far below leading-edge tech industries like software, IT
services, hardware, and networking, manufacturing ranks seventh out of
22 industries and scores higher than the median, or typical, industry.

In the executive summary of this book we note six findings from our
inquiry. In the customer stories we highlight a range of takeaways. In
the skills section we speak about the urgency to address the changing
workplace and workforce. And in the policy section we outline ethical
and regulatory considerations.

Throughout your reading of this book, we hope that a single overarching

theme emerges; one of optimism. Optimism that the inevitable march
of technological progress can, as it has for centuries, lead to better, safer,
and healthier lives when we work together to harness innovation for good
and plan for the unforeseen consequences with shared legal and ethical
principles.

We hope you will continue the discussion with your questions,

observations, and recommendations by joining us at:
https://news.microsoft.com/futurecomputed

119 120
FURTHER READING Charalambous, Eleftherios, Robert Feldmann, Gérard Richter, and
Christoph Schmitz. “AI in production: A game changer for manufacturers
Accenture (website). “Industry X.O.” Accessed March 2019. https://www.
with heavy assets.” McKinsey & Company, March 2019. https://www. FURTHER
READING

accenture.com/us-en/insights/industry-x-0-index.
mckinsey.com/business-functions/mckinsey-analytics/our-insights/ai-in-
Accenture Research. “Pivoting with AI.” Accenture, 2018. https://www. production-a-game-changer-for-manufacturers-with-heavy-assets.
accenture.com/t20180912T132343Z__w__/za-en/_acnmedia/PDF-85/
Chui, Michael and Sankalp Malhotra. “AI adoption advances, but
Accenture-Pivoting-with-AI-POV.pdf.
foundational barriers remain.” McKinsey & Company, November 2018.
Acemoglu, Daron and Pascual Restrepo. “Artificial Intelligence, https://www.mckinsey.com/featured-insights/artificial-intelligence/ai-
Automation, and Work.” In: The Economics of Artificial Intelligence: An adoption-advances-but-foundational-barriers-remain.
Agenda, ed. Ajay Agrawal, Joshua Gans, and Avi Goldfarb. Cambridge,
Coleman, Chris, Satish Damodaran, Mahesh Chandramouli, and Ed
MA: National Bureau of Economic Research, May 2019. p. 197-236.
Deuel. “Making Maintenance Smarter: Predictive Maintenance and the
https://www.nber.org/chapters/c14027.
Digital Supply Network.” Deloitte Insights, May 9, 2017.
Bauer, Harald, Peter Breuer, Gérard Richter, Jan Wüllenweber, Knut
https://www2.deloitte.com/insights/us/en/focus/industry-4-0/using-
Alicke, Matthias Breunig, Ondrej Burkacky, et al. “Smartening up with
predictive-technologies-for-asset-maintenance.html.
Artificial Intelligence (AI) – What’s in it for Germany and its Industrial
Sector?” McKinsey & Company, April 2017. https://www.mckinsey. Forbes Insights. “How AI Builds A Better Manufacturing Process.”
com/~/media/McKinsey/Industries/Semiconductors/Our%20Insights/ Forbes, July 17, 2018. https://www.forbes.com/sites/insights-
Smartening%20up%20with%20artificial%20intelligence/Smartening-up- intelai/2018/07/17/how-ai-builds-a-better-manufacturing-
with-artificial-intelligence.ashx. process/#28714b391e84.

Atluri, Venkat, Saloni Sahni, and Satya Rao. “The trillion-dollar Geissbauer, Reinhard, Evelyn Lübben, Stefan Schrauf, and Steve
opportunity for the industrial sector: How to extract full value from Pillsbury. Global Digital Operations 2018 Survey. PwC Strategy&, 2018.
technology.” McKinsey Digital, November 2018. https://www.mckinsey. https://www.strategyand.pwc.com/media/file/Global-Digital-Operations-
com/business-functions/digital-mckinsey/our-insights/the-trillion- Study_Digital-Champions.pdf.
dollar-opportunity-for-the-industrial-sector.
Giffi, Craig, Paul Wellener, Ben Dollar, Heather Ashton Manolian,
Castellina, Nicholas. “How Artificial Intelligence is Transforming the Luke Monck, and Chad Moutray. “2018 Skills Gap in Manufacturing
Manufacturing Workforce.” Manufacturing.net, September 2018. https:// Study.” Deloitte Insights, 2018. https://www2.deloitte.com/us/en/pages/
www.manufacturing.net/article/2018/09/how-artificial-intelligence- manufacturing/articles/future-of-manufacturing-skills-gap-study.html.
transforming-manufacturing-workforce.
Goering, Kevin, Richard Kelly, and Nick Mellors. “The Next Horizon
for Industrial Manufacturing: Adopting Disruptive Digital Technologies
in Making and Delivering.” McKinsey Digital, November 2018. https://
www.mckinsey.com/business-functions/digital-mckinsey/our-insights/
the-next-horizon-for-industrial-manufacturing.

121 122
Hawksworth, John and Yuval Fertig. “What Will Be the Net Impact of Perisic, Igor. “How Artificial Intelligence is Already Impacting Today’s
AI and Related Technologies on Jobs in China?” PwC, September 2018. Jobs.” Economic Graph (blog), LinkedIn, September 17, 2018.
https://www.pwc.com/gx/en/issues/artificial-intelligence/impact-of-ai- https://economicgraph.linkedin.com/blog/how-artificial-intelligence-is- FURTHER
READING

on-jobs-in-china.pdf. already-impacting-todays-jobs.

International Federation of Robotics. “Robot Density Rises Globally.” Philbeck, Thomas, Nicholas Davis, and Anne Marie Engtoft Larsen.
International Federation of Robotics press release, February 7, 2018. “Values, Ethics and Innovation Rethinking Technological Development
https://ifr.org/ifr-press-releases/news/robot-density-rises-globally. in the Fourth Industrial Revolution.” Cologny-Geneva: World Economic
Forum, January 2018. http://www3.weforum.org/docs/WEF_WP_Values_
Küpper, Daniel, Markus Lorenz, Kristian Kuhlmann, Olivier Bouffault,
Ethics_Innovation_2018.pdf.
Yew Heng Lim, Jonathan Van Wyck, Sebastian Köcher, et al. “AI in the
Factory of the Future: The Ghost in the Machine.” Boston Consulting PwC. “AI Will Create as Many Jobs as it Displaces by Boosting Economic
Group, April 2018. https://www.bcg.com/en-us/publications/2018/ Growth.” PwC press release, July 17, 2018. https://www.pwc.co.uk/press-
artificial-intelligence-factory-future.aspx. room/press-releases/AI-will-create-as-many-jobs-as-it-displaces-by-
boosting-economic-growth.html.
Kushmaro, Philip. “5 Ways Industrial AI is Revolutionizing
Manufacturing.” CIO, September 27, 2018. Rajadhyaksha, Ajay and Aroop Chatterjee. “Robots at the gate: Humans
https://www.cio.com/article/3309058/5-ways-industrial-ai-is- and technology at work.” Barclays, April 11, 2018. https://www.
revolutionizing-manufacturing.html. investmentbank.barclays.com/content/dam/barclaysmicrosites/ibpublic/
documents/our-insights/Robots-at-the-gate/Barclays-Impact-Series-3-
Microsoft Library ProResearch. “Artificial Intelligence (AI) & Ethics.”
Robots_at_the_Gate-3MB.pdf.
Topic paper, Microsoft Library ProResearch, June 11, 2018. https://
microsoft.sharepoint.com/sites/mslibrary/KeyTopics/Pages/Research/ Ransbotham, Sam, Philipp Gerbert, Martin Reeves, David Kiron, and
AI%20Ethics.pdf. Michael Spira. “Artificial Intelligence in Business Gets Real Pioneering
Companies Aim for AI at Scale.” MIT Sloan Management Review (Access
Mittal, Nitin, David Kuder, and Samir Hans. “AI-Fueled Organizations:
notes), September 2018. https://sloanreview.mit.edu/projects/artificial-
Reaching AI’s Full Potential in the Enterprise.” Deloitte Insights,
intelligence-in-business-gets-real/.
January 16, 2019. https://www2.deloitte.com/insights/us/en/focus/tech-
trends/2019/driving-ai-potential-organizations.html. Rao, Dr. Anand S. and Gerard Verweij. “Sizing the prize: What’s the real
value of AI for your business and how can you capitalise?” PwC, 2017.
Muro, Mark, Robert Maxim, and Jacob Whiton. Automation and
https://www.pwc.com/gx/en/issues/analytics/assets/pwc-ai-analysis-
Artificial Intelligence: How Machines are Affecting People and Place.
sizing-the-prize-report.pdf.
With contributions by Ian Hathaway. Washington, D.C.: Metropolitan
Policy Program at Brookings, January 24, 2019. https://www.brookings.
edu/wp-content/uploads/2019/01/2019.01_BrookingsMetro_Automation-
AI_Report_Muro-Maxim-Whiton-FINAL-version.pdf.

123 124
Rebello, Jagdish and Sachin Garg. Artificial Intelligence in Manufacturing Shum, Harry. “How to Ensure that We’re ‘Raising’ Ethical AI.” LinkedIn
Market – Global Forecast to 2023. With contributions by Tanuj Goyal, Pulse (blog), LinkedIn, May 17, 2018. https://www.linkedin.com/pulse/
Siddharth Rane, Anand Shanker, and Tarun Dutta. MarketsandMarkets, how-ensure-were-raising-ethical-ai-harry-shum/. FURTHER
READING

July 2017. https://microsoft.sharepoint.com/sites/mslibrary/research/

Somers, Ken. “Manufacturing’s Control Shift.” McKinsey & Company,
MktResearch/MarketsAndMarkets/Artificial_Intelligence_in_
August 22, 2018. https://www.mckinsey.com/business-functions/
Manufacturing_Market_Global_Forecast.pdf#search=artificial%20
operations/our-insights/operations-blog/manufacturings-control-shift.
intelligence%20manufacturing.
The Manufacturer. “Annual Manufacturing Report 2018.” London:
Rodriguez, Michelle Drew, Robert Libbey, Sandeepan Mondal,
Hennik Research, May 2018. https://cdn2.hubspot.net/hubfs/728233/
Jeff Carbeck, and Joann Michalik. “Exponential Technologies in
AMR-2018-Amended-May-2018.pdf.
Manufacturing.” Deloitte, 2018. https://www2.deloitte.com/us/en/pages/
manufacturing/articles/advanced-manufacturing-technologies-report. Tratz-Ryan, Bettina, Alexander Hoeppe, and Pete Basiliere. “Predicts
html. 2018: Industries 4.0 and Advanced Manufacturing.” Gartner (Access
notes), April 9, 2018. https://www.gartner.com/document/3871170.
Rose, Justin, Vladimir Lukic, Claudio Knizek, Tom Milton, Alex Melecki,
and Howie Choset. “Advancing Robotics to Boost US Manufacturing Tratz-Ryan, Bettina and Michelle Duerst. Manufacturing Industries
Competitiveness.” Boston Consulting Group, October 25, 2018. https:// Digitalization Primer for 2019. Gartner (Access notes), February 5, 2019.
www.bcg.com/en-us/publications/2018/advancing-robotics-boost-us- https://www.gartner.com/document/3899977.
manufacturing-competitiveness.aspx.
Wellener, Paul, Ben Dollar, and Heather Ashton Manolian. “The Future
Schaeffer, Eric, Jean Cabanes, and Abhishek Gupta. “Manufacturing of Work in Manufacturing.” Deloitte Insights, March 21, 2019. https://
the Future.” Accenture, 2018. https://www.accenture.com/ www2.deloitte.com/insights/us/en/industry/manufacturing/future-of-
t20180327T080053Z__w__/us-en/_acnmedia/PDF-74/Accenture-Pov- work-manufacturing-jobs-in-digital-era.html.
Manufacturing-Digital-Final.pdf#zoom=50.
World Economic Forum and McKinsey & Company. “Fourth Industrial
Schaeffer, Eric, Raghav M. Narsalay, Oliver Hobräck, Matthias Revolution: Beacons of Technology and Innovation in Manufacturing.”
Wahrendorff, and Abhishek Gupta. “Turning Possibility into Cologny-Geneva: World Economic Forum, January 2019.
Productivity.” Accenture, 2018. https://www.accenture.com/
http://www3.weforum.org/docs/WEF_4IR_Beacons_of_Technology_and_
t20180516T150528Z__w__/us-en/_acnmedia/PDF-78/Accenture-
Innovation_in_Manufacturing_report_2019.pdf.
IndustryX0-AI-products_RD.pdf#zoom=50.

Shaw, Jonathan. “Artificial Intelligence and Ethics.” Harvard Magazine,

January-February 2019. https://harvardmagazine.com/2019/01/artificial-
intelligence-limitations.

125 126
World Economic Forum and McKinsey & Company. “The Next
Economic Growth Engine: Scaling Fourth Industrial Revolution
Technologies in Production.” Cologny-Geneva: World Economic Forum, FURTHER
READING

January 2018. http://www3.weforum.org/docs/WEF_Technology_and_

Innovation_The_Next_Economic_Growth_Engine.pdf.

Zhong, Ray Y., Xun Xu, Eberhard Klotz, and Stephen T. Newman.
“Intelligent Manufacturing in the Context of Industry 4.0: A Review.”
Engineering vol. 3, no. 5 (October 2017): 616-630. https://doi.
org/10.1016/J.ENG.2017.05.015.

127 128
ENDNOTES 8 Gates, Dominic. “Boeing invests in advanced supersonic business
jet.” Seattle Times, Feb. 5, 2019. https://www.seattletimes.com/business/
1 Long, Katie, Donna Tam, Carrie Barber, Arjuna Soriano, and boeing-aerospace/boeing-invests-in-advanced-supersonic-business-jet/ ENDNOTES

Paul Brent. “Robot-Proof Jobs.” Marketplace, 2017. https://features.

9 “Will AI Destroy More Jobs Than It Creates Over the Next
marketplace.org/robotproof/
Decade?” Wall Street Journal, April 1, 2019. https://www.wsj.com/
2 Prodhan, Georgina. “European parliament calls for robot law, articles/will-ai-destroy-more-jobs-than-it-creates-over-the-next-
rejects robot tax.” Reuters Technology News, Feb. 16, 2017. https://www. decade-11554156299
reuters.com/article/us-europe-robots-lawmaking-idUSKBN15V2KM
10 Greenhouse, Steven. “Unions Face The Fight Of Their Lives To
3 Giffi, Craig A., Paul Wellener, Ben Dollar, Heather Ashton Protect American Workers.” Huffington Post, May 4, 2018. https://www.
Manolian, Luke Monck, and Chad Moutray. “2018 skills gap in huffpost.com/entry/american-workers-jobs-inequality-union-automation
manufacturing study, Future of manufacturing: The jobs are here, but _n_5ae043f9e4b061c0bfa32e0c
where are the people?” Deloitte, Nov. 14, 2018. https://www2.deloitte.
11 “UNI Global Union calls for the establishment of a global
com/us/en/pages/manufacturing/articles/future-of-manufacturing-skills-
convention on ethical artificial intelligence.” UNI Global Union, Dec.
gap-study.html
14, 2016. https://www.uniglobalunion.org/news/uni-global-union-calls-
4 “Which Roles are Most Difficult to Fill in the UK?” 2018 establishment-a-global-convention-ethical-artificial-intelligence
Talent Shortage Survey. Manpower Group, Feb. 7, 2018. https://www.
12 “10 Principles for Ethical AI.” UNI Global Union. Accessed May
manpowergroup.co.uk/the-word-on-work/2018-talent-shortage-
19, 2019. http://www.thefutureworldofwork.org/opinions/10-principles-
survey/#hard-to-fill-roles
for-ethical-ai/
5 Rao, Dr. Anand S., and Gerard Verweij. Sizing the prize: What’s
13 Meeting of the OECD Global Parliamentary Network, February
the real value of AI for your business and how can you capitalise?
13-15, 2019. http://www.oecd.org/parliamentarians/meetings/gpn-
PwC, 2017. https://www.pwc.com/gx/en/issues/analytics/assets/pwc-ai-
meeting-february-2019/
analysis-sizing-the-prize-report.pdf
14 “2017 Survey Of Occupational Injuries & Illnesses: Charts
6 Purdy, Mark, and Paul Daugherty. “Future of Artificial
Package.” U.S. Bureau of Labor Statistics. U.S. Department of Labor, Nov.
Intelligence: Artificial Intelligence Is The Future Of Growth.” Accenture
8, 2018. https://www.bls.gov/iif/osch0062.pdf
LLP. Accessed May 19, 2019. https://www.accenture.com/us-en/insight-
artificial-intelligence-future-growth 15 LinkedIn Workforce Report. Accessed May 19, 2019. https://
economicgraph.linkedin.com/resources#view-all
7 “The Federal Government’s Artificial Intelligence Strategy: ‘AI
Made in Germany.” German Federal Ministry for Economic Affairs 16 Perisic, Igor. “How artificial intelligence is already impacting
and Energy. Accessed May 19, 2019. https://www.de.digital/DIGITAL/ today’s jobs.” LinkedIn Economic Graph, Sept. 17, 2018. https://
Redaktion/EN/Standardartikel/artificial-intelligence-strategy.html economicgraph.linkedin.com/blog/how-artificial-intelligence-is-already-
impacting-todays-jobs

129 130
17 “The skills gap in U.S. manufacturing 2015 25 “Adobe, Microsoft and SAP announce the Open Data Initiative
and beyond.” Manufacturing Institute and Deloitte, to empower a new generation of customer experiences.” Microsoft News
2015. http://www.themanufacturinginstitute.org/~/ Center, Sept. 24, 2018. https://news.microsoft.com/2018/09/24/adobe- ENDNOTES

media/827DBC76533942679A15EF7067A704CD.ashx microsoft-and-sap-announce-the-open-data-initiative-to-empower-a-
new-generation-of-customer-experiences/
18 “Manufacturing Our Future: Cases on the Future of
Manufacturing.” World Economic Forum, May 2016. http://www3. 26 “Manufacturers with artificial intelligence to nearly double
weforum.org/docs/GAC16_The_Future_of_Manufacturing_report.pdf competitiveness.” Microsoft Asia News Center, April 1, 2019. https://
news.microsoft.com/apac/2019/04/01/manufacturers-with-artificial-
19 Boyle, Alan. “Microsoft CEO Satya Nadella lays out 10 Laws of AI
intelligence-to-nearly-double-competitiveness/
(and Human Behavior).” Geekwire, June 28, 2016. https://www.geekwire.
com/2016/microsoft-ceo-satya-nadella-10-laws-ai/ 27 Partnership on AI. Accessed May 19, 2019. https://www.
partnershiponai.org/
20 Wang, Amy B. “‘I’m in your baby’s room’: A hacker took over a
baby monitor and broadcast threats, parents say.” Washington Post, Dec. 28 “National And International AI Strategies.” Future of Life
20, 2018. https://www.washingtonpost.com/technology/2018/12/20/nest- Institute. Accessed May 19, 2019. https://futureoflife.org/national-
cam-baby-monitor-hacked-kidnap-threat-came-device-parents-say/ international-ai-strategies/?cn-reloaded=1

21 Content of Premarket Submissions for Management of 29 Accelerating Sustainable Production. World Economic Forum.
Cybersecurity in Medical Devices: Draft Guidance for Industry and Food Accessed May 19, 2019. https://www.weforum.org/projects/accelerating-
and Drug Administration Staff. U.S. Food and Drug Administration, Oct. sustainable-production
18, 2018. https://www.fda.gov/media/119933/download
30 “The Hidden Water In Everyday Products.” Water Footprint
22 “Age Demographics for Industry Workforces.” Governing. Calculator, July 1, 2017. https://www.watercalculator.org/water-use/the-
Accessed May 19, 2019. https://www.governing.com/gov-data/ages-of- hidden-water-in-everyday-products/
workforce-for-industries-average-medians.html
31 “The Hidden Water In Everyday Products.” FirstPost, April 22,
23 Romei, Valentina. “How Japan’s ageing population is 2019. https://www.firstpost.com/tech/news-analysis/earth-day-2019-
shrinking GDP.” Financial Times, May 16, 2018. https://www.ft.com/ ai-has-a-huge-role-to-play-in-fulfilling-sustainable-development-
content/7ce47bd0-545f-11e8-b3ee-41e0209208ec goals-6486441.html

24 “Shaw, Greg. The Ability Hacks. Microsoft, 2018. https:// 32 “Global Warming of 1.5 ºC.” Intergovernmental Panel on Climate
blogs.microsoft.com/wp-content/uploads/prod/sites/5/2018/08/ Change. Accessed May 19, 2019. https://www.ipcc.ch/sr15/
theabilityhacksbook.pdf

131 132
33 United Nations Sustainable Development Goals. United Nations.
Accessed May 19, 2019. https://sustainabledevelopment.un.org/sdgs. See
UN Sustainable Development Goals 12 and 13. ENDNOTES

34 “Sustainability’s Strategic Worth.” McKinsey & Co., July 2014.

https://www.mckinsey.com/business-functions/sustainability/our-
insights/sustainabilitys-strategic-worth-mckinsey-global-survey-results

35 Banham, Russ. “Industry 2050: How Clean Manufacturing

Is A Win-Win Proposition.” Mitsubishi Heavy Industries
Brandvoice. Forbes, Oct. 18, 2018, https://www.forbes.com/sites/
mitsubishiheavyindustries/2018/10/18/industry-2050-how-clean-
manufacturing-is-a-win-win-proposition/#473a13b17f35

36 Herweijer, Celine, Benjamin Combes, Jonathan Gillham, Lucas

Joppa, et al. “How AI Can Enable a Sustainable Future.” Microsoft and
PwC. Accessed May 19, 2019. https://www.pwc.co.uk/sustainability-
climate-change/assets/pdf/how-ai-can-enable-a-sustainable-future.pdf

37 “How AI can enable a sustainable future: Estimating the

economic and emissions impact of AI adoption in agriculture, water,
energy and transport.” PwC UK. Accessed May 19, 2019. https://www.
pwc.co.uk/services/sustainability-climate-change/insights/how-ai-
future-can-enable-sustainable-future.html

38 “OECD Sustainable Manufacturing Toolkit: Seven Steps to

Environmental Excellence.” Organisation for Economic Co-operation and
Development, 2011. http://www.oecd.org/innovation/green/toolkit/#d.
en.192438

39 http://www.oecd.org/innovation/green/toolkit/
aboutsustainablemanufacturingandthetoolkit.htm

133 134
ACKNOWLEDGEMENTS Greg Shaw is Senior Director in the Office of the CEO at Microsoft
and co-author of Hit Refresh by Satya Nadella (Harper Collins)
and The Ability Hacks. Across a long career at Microsoft, Greg
has served in many senior communications and community facing
roles, including as a writer for Bill Gates and helping create the
I would like to thank the following contributors for providing their company’s giving program to provide access to computing and the
insights and perspectives in the development of this book. Internet through public libraries.

Barbara Olagaray Gatto, Guy Berger, Hemant Pathak, Indranil Sircar Jane Prior to Microsoft, Greg was publisher and CEO of Crosscut.com,
Broom Davidson, Jack Chen, Jeremy Rollison, Karon Kocher, Lucas Joppa, an online news and opinion magazine. He also served for nearly
Marcus Bartley Johns, Mark Lange, Mike Phillips, Nick Tsilas, a decade as an executive at the Bill & Melinda Gates Foundation,
Owen Larter, Portia Wu, Stephanie Rowland, Steve Guggenheimer, overseeing policy and advocacy for the foundation’s U.S. Program.
Steve Sweetman, Thomas Roca, Tracy Kennedy
Greg has been appointed to numerous start-up, government
And special thanks to John Galligan, our series editor. and nonprofit boards, and is asked frequently to be a speaker on
news, media and publishing. He has a B.A. in journalism from
Northeastern State University in Oklahoma. Greg and his wife
have two children.

GREG SHAW

135 136
137 138
139