The Business Case For AI in HR
The Business Case For AI in HR
The Business Case For AI in HR
for AI in HR
Insights and tips on getting started
Foreword Introduction
HR is on the brink of massive disruption. The ubiquity of mobile Every aspect of life and work is being transformed by AI.
and social technologies and personalization has redefined the Leading organizations understand the impact of AI on
bar for employee experience. With the disruption of business business models, workforce demographics, and the changing
models across all industries, the talent acquisition game has experiences expected by customers and employees alike.
changed and with the impact of technological change on work, AI combined with strategic insight creates new business
the need for a continuous learning culture has never been more opportunity and is transforming the way HR contributes to an
urgent. The CHRO stands at the intersection of all these organization’s competitive advantage.
demands, generally with a declining budget and ongoing
operational distractions. This report outlines tangible examples of where AI is delivering
value in HR today. It describes the pioneering work of IBM’s
Just at this opportune moment, AI and automation are own internal HR team, who together with IBM’s client services
becoming accessible for the HR profession. In my experience, experts have developed AI solutions for HR that are truly
it’s enabling us to solve pervasive talent issues such as knowing remarkable. Many of the examples described, which were
our skills, preventing unwanted employee turnover, reacting developed initially for internal IBM employee use, delivered
quickly to employee hot spots, matching employees and such significant value that they are now offered commercially.
external candidates with career opportunities, supporting These include IBM Watson Candidate Assistant, IBM Watson
managers with better salary investment guidance, eliminating Recruitment, IBM Watson Career Coach, and Your Learning.
manual tasks in benefits administration and payroll through
robotic process automation, and creating an irresistible For the last decade, IBM has been proud to work with clients
platform for employees to learn on the go. around the world on their most important transformations. We
help our CHRO clients and their teams on their HR reinvention
With some upskilling, ethical operating guidelines, and a paths, building the business cases for investment, ensuring
healthy dose of technical curiosity, the HR function is now meaningful business and employee outcomes, and providing
positioned to truly drive strategic advantage while better unique insights into how to manage change driven by digital
supporting the workforce we rely on to put the strategy and AI reinvention. As a leading talent-centric, AI-powered
into action. organization, we are excited to be at the forefront of
this journey.
This report tells the story of what’s possible and how to get
there, with specific examples showcasing what we’ve done Tina Marron-Partridge
at IBM, and practical tips to help others embark on their Global Leader, Talent & Transformation, IBM
AI journey in HR. And, as I witness the benefits first hand,
I couldn’t be more optimistic about the possibilities that
lie ahead.
Diane Gherson
Chief Human Resource Officer, IBM
Page 2
Contents Executive summary
4 Executive Summary IBM’s HR function was one of the first to adopt artificial
intelligence (AI) technology and this means it has a wealth of
4 AI signals an HR technology step change insights and learnings to share to help others get started. In this
5 Why AI is being used in HR report, the IBM Smarter Workforce Institute summarizes those
learnings, which were gathered by interviewing the senior
5 An HR ‘moon shot’ human resource executives responsible for bringing AI to the
6 How AI can be used in HR HR function at IBM.
– Attract: Enhancing candidate experience
The executive interviews revealed AI is effective in HR because
– Hire: Efficient and effective recruitment
it helps:
– Engage: Enhancing motivation
– Retain: Smarter compensation planning – Solve business challenges
– Develop: Personalized learning – Attract and develop new skills
– Grow: Career development – Improve employee experiences
– Serve: AI for 24/7 employee interaction – Provide analytical decision support
11 The benefits of AI in HR – Make more efficient use of HR budgets
– Return on investment
– Benefits and outcome metrics IBM HR’s experience is that AI can be applied in almost any
– Time to results area of HR, including candidate attraction, hiring, learning,
compensation, career management, and HR support. This
13 Five steps to getting started report includes use cases across the employee journey. In each
– Step 1: Start with a business case area, we describe some of the benefits IBM has seen since
– Step 2: Decide whether to buy or build implementing AI. We also cover practical topics such as how to
– Step 3: Identify the skills you have and need get started, the skills you will need, and important issues about
– Step 4: Implement MVP fairness and the broader societal effects of AI on jobs.
– Step 5: Roll out enterprise-wise
18 Acknowledgements
Page 3
AI signals an HR technology step change “AI is an accelerator—it allows us the ability to
HR departments were once primarily administrative functions. ingest a variety of data and provide context to a
Referred to as personnel departments, the key responsibilities decision maker or employee or business leader.
were clerical, and work in the HR department focused on record
keeping about the workforce. But the view of human resources It allows us to deliver the right intelligence in the
has evolved considerably in the last 30 years. Research has moment and achieve personalization at scale.”
shown the ways organizations manage their workers have
important implications for how well organizations perform.1 – Tom Stachura, Vice President Talent Solutions
& People Analytics, IBM
Today the phrase ‘strategic HR’ is used to refer to HR practices
that provide a competitive advantage to organizations.2 The
Why AI is being used in HR
strategic HR movement has seen a shift in HR’s focus from
administrative practices to high performance HR practices like Today, AI’s capabilities are being used to augment business
teamwork and performance management, which focus on key operations and consumer solutions. We have identified five
jobs rather than every job and on groups of critical workers primary reasons for implementing AI in HR:
rather than every worker.3
– To solve pressing business challenges
Until recently, the primary benefit of technology has been to AI enables HR organizations to deliver new insights and
provide efficiency gains; it allowed us to do the same things we services at scale without ballooning headcount or cost.
always did, but faster and more cost effectively.4 For example, Persistent challenges, like having the people resources to
previously technology allowed us to recruit people faster over deliver on the business strategy and allocating financial
the internet, but now AI lets us recruit the right people faster resources accordingly, can be addressed through the
by assessing skill match for roles, predicting the likelihood thoughtful application of AI solutions.
of future success, and estimating the expected time to fill
– To attract and develop new skills
any given role. This is an example of the ways in which AI
The business world is constantly being disrupted. In order
is changing the situation so that technology enables the HR
to cope with this disruption, businesses need to respond
function to solve critical business challenges, building on earlier
faster to opportunities, and to work in an agile way to stay
contributions from workforce analytics. Where previous HR
ahead of competitors. This means finding an effective way
initiatives led to incremental change, AI offers the opportunity
to compete for the skills required to innovate in this new
for exponential performance improvements in HR.
operating environment. AI applications enable HR
departments to acquire and develop employee skills
in lockstep with shifting market demand.
Page 4
To use HR budgets as efficiently as possible Research methodology
AI can enable HR to become more efficient with its funding.
For this report, IBM Smarter Workforce Institute conducted
HR spend can shift to higher value and more complex problem
20 60-minute in-depth structured interviews with the most
solving, without reducing levels of service for workers who
senior HR executives responsible for bringing AI to HR at IBM.
have more routine HR queries. HR savings made in this way
Interview participants included experts in talent acquisition,
can be reinvested in further AI deployment, increasing HR’s
testing and selection, learning and development, talent
ability to solve business challenges, continuously develop
management, compensation and benefits, performance
strategic skills, create positive work experiences, and provide
management, engagement and culture, employee and labor
outstanding decision support for employees.
relations, computer science, analytics, HR technology, and
general HR.
An HR ‘moon shot’
As you read through this paper, we hope you’ll keep in mind Interview topics included:
the idea of a ‘moon shot’ for applications of AI in HR. Like
– Objectives of AI in HR and types of projects
President John F. Kennedy’s 1961 seemingly impossible goal
– Funding, timelines, and benefits
of sending a person to the moon within 10 years when the
– Skills needed
technology for such a feat did not exist, AI opens the door to
– Impact on jobs
previously unimaginable possibilities. For example, a potential
– Bias, diversity, and inclusion
moon shot for driverless cars is the objective of zero accident-
– Tips for getting started
related motor vehicle deaths. Applications of AI in healthcare
have the ultimate objective of eradicating diseases. What
Interview responses were analyzed for common themes and
might the moon shot be for applications of AI in HR? A moon
key insights, as summarized throughout this paper.
shot for AI in HR could be that employees are in complete
control of their careers because AI helps their skills evolve at
the same speed that technology evolves. In other words, a “Ask yourself: what things would be better if
moon shot could be that AI helps workers renew their skills they were done 24/7? What would be better if
before existing skills become obsolete. Different organizations
will likely have different moon shots for AI in HR. What is the it were done at scale? What would benefit from
moonshot in your organization? greater consistency? What would be possible if
we leveraged broader expertise to see beyond
our current limits? These are good candidates
for AI.”
– Debora Bubb, Vice President and Chief Leadership,
Learning & Inclusion Officer, IBM
Page 5
How AI can be used in HR “When we piloted AI for candidate attraction we
Deployment of AI in HR can occur across the entire talent saw a big increase in candidates applying for
lifecycle (see Figure 1). In the next section, examples of AI jobs at IBM, and there was greater stickiness.
use cases at each point in the talent lifecycle are outlined.
In addition, Net Promoter Score feedback said
the AI was engaging. People felt it answered
Serve their questions, was relevant and useful.”
– Joanna Daly, Vice President Talent, IBM
Page 6
Hire: Efficient and effective recruitment AI manager alerts at IBM
At IBM, managers get alerts tailored to the needs of individual
The job of a recruiter is time pressured and complex, often
employees. For example, if someone has been on a team
having to fill many roles at once. Recruiters need to prioritize
a long time, has certain skills, and is ready for a promotion,
all of the different roles they are responsible for, and at the
the manager is alerted to these facts. Similarly, managers
same time, they need a way to differentiate among candidates
receive alerts about employees with a greater propensity to
competing for the same role. Not meeting these challenges
leave. When sales people are at risk of missing quotas, early
effectively enough can mean the wrong roles get prioritized,
interventions can be suggested to get people back on track.
and even where the right roles are prioritized, the wrong
Alerts such as these enable managers to make decisions that
candidates might be selected for roles. AI can be used in this
are consistent with an organization’s talent management
setting to predict how long a job requisition will take to fill
approach by recommending decisions that HR would like to
based on historical data, allowing recruiters to reprioritize as
see implemented.
needed. AI can also be used to determine the match between
a candidate’s resume and the job requisition, and to make AI chatter analysis at IBM
accurate predictions of future performance based on At IBM, chatter analysis is used to surface the top three
information about the candidate collected in the job issues from social media sources within the company firewall.
application process. Furthermore, it can help recruiters write This provides recommendations that are personalized to a
more inclusive job descriptions and filter candidates more specific leader to help improve engagement in their team.
effectively, minimizing the impact of unconscious bias in their If an employee receives recognition for outstanding work, for
process and practices. example, IBM might recommend to their manager to amplify
that feedback by sharing it with others. IBM has observed that
AI recruitment at IBM
these sorts of actions improve engagement. ‘Engage at IBM’
In a large organization like IBM, effective prioritization
is an AI application that learns; the leader provides feedback
of recruitment demands careful selection of applicants.
on the recommendations and the system will improve as a
IBM needed a better way to help recruiters surface the top
result. As the system gets better, so too does the managers’
candidates for open jobs and to prioritize the most important
effectiveness at managing and inspiring their teams.
requisitions. The solution developed, IBM Watson Recruitment
(IWR), uses AI to leverage information about the job market and
past experiences of hiring candidates to predict time to fill and
identify the candidates most likely to be successful.
Page 7
“If compensation decisions are based on just Develop: Personalized learning
one or two data points, such as tenure and AI in a learning context can help accelerate skill development
performance, a manager can make the decision at the level of the individual, and it can optimize learning at
the level of the organization. One of the areas with the most
without analytical support. But managers promise is AI tagging of learning content. In the past, when
should consider many factors, such as market learners interacted with a learning management system, the
rates and propensity to learn. With more data content they found depended largely on the descriptions that
the developers uploaded with the training in the first place.
points, AI is needed to avoid underpaying some These descriptions are called metadata. With AI tagging,
and overpaying others.” resources such as images and documents that are uploaded
to learning management systems are enriched with metadata
– Nickle LaMoreaux, Vice President Compensation through AI technologies, and this helps learners locate the
& Benefits, IBM training content and use it again and again more efficiently.
Making sound compensation decisions requires careful – Open learning platform: Integrates employee and learning
consideration of a wide range of factors. In addition to data from various sources, bringing all relevant content
performance, these factors include the market rate for the together for access from any device—allowing learning
skills, how in demand the skills are, and whether it is better to to happen anywhere and at any time.
reward strong performance in base pay or in bonuses. To make – Employee-specific experience: Provides personalized
optimal base pay decisions that reflect this line of thinking, you learning recommendations tailored to job role, business
need a deep understanding of employee skills, the going rate group, skill set, and personal learning history—encouraging
for those skills, and whether those skills are increasing or continuous employee development and skill growth.
decreasing in demand. There can be many more data points – Content channels: Learning content is organized to
that need to be considered than a person can analyze without support a variety of needs and interests—resulting in simpler
analytical support. The advent of AI-based compensation browsing and ongoing development aligned
support can lead to thousands of hours of preparation for with business initiatives.
compensation cycles being reduced to just a few hours, while
providing decision advice that examines many more variables By making learning easily available when and where it
than were previously considered. Furthermore, by focusing on is needed, AI helps the acquisition of strategic skills for
skills in determining compensation, the use of AI minimizes organizations. AI can be used to create an overall picture
chances that bias exists in the compensation process. of how the organization is doing in the area of learning in
the form of a learning dashboard. Dashboards can show
AI-supported compensation planning at IBM
progress towards closing identified skill gaps in a business.
Making complex compensation decisions accurately across
The learning history of particular individuals in an organization
an organization is a challenge, and one that IBM uses AI to
can even be considered an indication of a person’s propensity
address. IBM designed an AI-powered decision support tool
to learn. Propensity to learn is becoming just as important as a
that assists with compensation planning, helping managers
person’s current skills as the shelf-life of skills continues to
avoid underweighting or overweighting the critical data points.
decrease at a fast rate.
The application reviews dozens of data points in making its
recommendations, integrating external information from
sources like the Bureau of Labor Statistics with internal data “With AI, we are able to see how learning
on factors such as cost to replace. The application is currently relates to engagement without even having
being deployed for tens of thousands of first-line managers to
to ask our employees for their perceptions.
assist with their compensation planning, following successful
early trials in focused geographies. If managers went to manager training, are
their people more engaged? We have the
Importantly, when using the tool, managers have the
opportunity to override the AI recommendation about any answer and that gives us important feedback
given employee, and the system can continue to learn from on how effective our training is.”
managers’ actual decisions. In general, managers tend to
follow the recommendations the AI provides, and this has – Gordon Fuller, Vice President and Chief Learning Officer, IBM
helped ensure employees are not overpaid or underpaid
at IBM. IBM also emphasizes transparency in AI-based
compensation support: employees can see where they sit
relative to the market, because the low and high range of
compensation for workers with their skills is provided, in
addition to their personal salary.
Page 8
AI-based learning at IBM
At IBM, the introduction of AI in learning has produced
impressive results. Using AI, IBM has developed Your Learning,
a rich, personalized digital marketplace for learning, visited by
98% of employees every quarter. IBMers complete an average
of 60 hours of learning per year. IBMers can navigate to the
learning most popular with their peers, sign up for targeted
learning channels, and explore the skills and badges they need
to prepare for the company’s hottest roles. A learning chatbot
answers questions 24/7. As a result, IBM’s AI-driven learning
platform is increasing enrollments and course completions,
thereby accelerating strategic skill acquisition.
Page 9
Grow: Career development Chatbots are inexpensive to design, and quick to train. Chatbot
implementation can be managed even by non-technical HR
Career coaching is a powerful way to create more meaningful
staff because of the wide availability of easy-to-use application
work experiences for employees and can lead to greater
programming interfaces (APIs). For these reasons, chatbots are
productivity and stronger succession pipelines for businesses.
a natural starting point for organizations wanting to build AI into
Career coaching is usually high touch, time intensive, and
their HR function. If you have an HR process that generates a
expensive. This has meant that coaching has usually been
lot of questions from employees, you can deploy a chatbot
reserved for people who were either underperforming, or for
based on a well-developed frequently asked questions (FAQ)
high potential individuals that an organization was prepared to
list about the HR process.
invest a considerable amount of money in. However, bringing
this level of opportunity for employee career development to Chatbots are a perfect example of where AI can improve the
everyone in the organization is likely to lead to benefits to both employee experience, because they provide real-time answers
the employees whose motivation and engagement increases, at any time, day or night. Chatbots also constantly learn from
and also the organization itself through improved employee feedback, and improve the answers given.
job performance overall.
It is always a good idea to involve people with communications
AI career advice at IBM experience in the design of the chatbot. Communications
IBM wanted to provide high quality career guidance to all experts can help work on word choices and tone of the bot
employees, regardless of level. The solution developed is a to increase personalization.
personal advisor, Watson Career Coach (WCC), that brings the
benefits of career coaching to all of IBM. The personal advisor Nirvana with chatbots will be bots that are embedded in the
is an AI assistant that interacts with employees who are workflow of employees and help with work rather than just
thinking about future opportunities. It gets to know the answer questions about how to do the work. For example, if a
employee by asking and answering questions with natural manager needs to transfer an employee to a new department,
language and integrating with historical information. the interaction with the bot would result in actually executing
the transfer. This is referred to as ‘conversational HR,’ which
A second component of the solution, job opportunity match, extends the functionality of the bot.
allows employees to upload their resumes or answer
questions about skills, and then suggests roles that are AI chatbot use in IBM HR
suitable for the employee. If workers are looking for IBM has chatbots that are used by employees all year round in
opportunities further out, they can use a career navigator areas across HR. For example, HR has deployed chatbots to
component. The career navigator lets workers plan their way support employees with their benefits enrollment decisions,
towards desired roles and prepares them for growth towards and to support managers with their compensation planning;
those roles with development suggestions that will help them both are areas with designated time periods characterized by
build the required skills. The AI assistant can also offer high usage rates, requiring fast responses to user questions.
personalized learning recommendations which help
employees continuously develop their skills. Chatbots that are busier at certain times of the year, such
as the performance management, benefits enrollment, and
compensation planning chatbots, are considered ‘seasonal
Serve: AI for 24/7 employee interaction
bots.’ IBM also has bots that are accessed 24 hours a day,
A chatbot is an application that understands natural language seven days a week, year-round. An example is IBM’s popular
and interacts with users in human-like conversation. Chatbots new-hire chatbot. It is one of the busiest chatbots at IBM,
are arguably the AI technology with the broadest application answering 700 questions a day. New hire chatbots are
to HR problems, and they have been deployed in most areas particularly helpful because they resolve the challenge
of HR. of not knowing who to ask for help.
Page 10
The benefits of AI in HR
Financial Metrics
As seen in the use cases described in this report, AI in HR has
e.g., influenced revenue
the potential for significant organizational benefits. This was
achieved by clear articulation of expected outcomes tied to
business problems, careful selection of appropriate metrics to
measure the expected outcomes, and regular tracking of results
to enable iterative improvements. HR Metrics
e.g., skill breadth and depth
Return on investment
Tracking return on investment (ROI) in HR needs to be a
business imperative. HR practitioners should have a direct line
of sight from AI applications to the outcomes AI will produce Outcomes Produced by AI
and the associated ROI that occurs in the business. Establishing e.g., personalized learning
the expected connection between AI and its return should
occur before the AI application is implemented. A framework
for thinking about how AI ROI can be examined is outlined in
Figure 2. A framework for structuring an AI business case
Figure 2. This framework shows that results produced by AI
should have a clear connection to HR metrics, and in turn, HR
metrics should have a clear connection to financial metrics. Note: The business case does not need to be complex. You should aim
to make the case that AI applications lead to desired outcomes, which
Let’s consider an example of an organization struggling to close in turn influence HR metrics, which then impact financial metrics.
sales because of a lack of technical knowledge among its sales
people. Here, AI might be proposed as a method to personalize
learning for the individuals’ needs, delivered in a consumable
format (e.g., mobile). Personalized learning is the outcome that Demonstrating return on AI investment at IBM
the AI produces. The HR metric that the learning connects to At IBM, all proposals for building an AI application in HR
is the skill breadth and depth of sales people in the business. require a business case. Once the AI applications are running
The AI system should track who did and did not complete in the business, there is a rigorous, quarterly management
their personalized learning, and the organization should track system to track the HR, financial and NPS metrics. There
whether these individuals were associated with successful are also metrics tracked that are specific to particular AI
sales deals. Dollar values can be put on the sales deals that applications. With its candidate attraction AI applications,
were generated, and the difference in revenue between the IBM has seen sizable increases in candidates applying for jobs.
groups, relative to the cost of the AI, can be examined across IBM’s recruitment AI application provides the capability to
two groups: those who completed the AI based training and efficiently and effectively process the increased volume and
those who did not. If the difference is estimated before the AI has led to faster time to hire candidates as well as better quality
system is implemented, the benefit is a projection and should candidates. Overall, in HR in 2017 alone, IBM realized $107
form part of the business case. If the figure is an actual value million in savings as a result of AI.
calculated after the implementation of the system, the figure
Benefits and outcome metrics
represents a check on the ROI of the application.
Figure 3 provides examples of expected benefits and outcome
It is impossible to show unequivocally that ROI observed metrics for each of the AI applications discussed. Look for a
is strictly attributable to the application. But careful designs combination of near term and longer term metrics so that you
(e.g. pre and post implementation comparisons, use of control can evaluate quickly and course correct as needed.
groups) can reveal converging lines of evidence pointing
towards ROI. For further discussion of designs that can be
used please see the book, The Power of People: Learn How
Successful Organizations Use Workforce Analytics To Improve
Business Performance5, or the white paper, Decoding
Workforce Analytics6.
Page 11
Application of AI in HR Examples of expected benefits Examples of outcome measures
Efficient and effective – Better prioritization of job requisitions – Skill shortages or unfilled vacancies
recruitment – Accelerated time-to-hire – Average time to fill open positions
– Accurate assessment of diverse candidates – Selection ratios of minority and
– Identification of the most qualified candidates majority candidates
– New hire productivity
Enhanced motivation – Better manager support for their employees – Employee retention
– Improved employee experience – Engagement or experience survey scores
Smarter compensation – Increased pay transparency for employees – Compensation satisfaction survey scores
planning – Optimized compensation budgets aligned – Overpaid or underpaid worker count
with business strategy
Personalized learning – Accelerated employee skill acquisition – Match between current and required skills
– Better alignment of employee skills with mix, skill gap closure
business strategy – Employee productivity
– Enhanced learning experience – Course enrollments and completion rates
Career development – Employee driven career management – Career satisfaction survey scores
for all – Employee clarity on opportunities – Number of internal job applications
and moves
24/7/365 Employee – Better informed and more productive – Number of process violations or exceptions
support employees via faster, more accurate – Labor costs
answers to questions
– Reduced number of support center staff
Page 12
Time to results “Think about the process or solution in your
The timeline to results depends on two factors, the decision to
buy or build and the nature of the AI project being undertaken.
area of responsibility that provides the most
challenge. What is it that you need the most
Whether you buy or build
help with, and what from an AI perspective will
If you buy off-the-shelf AI solutions, you can expect to see
results in term of ROI within the first three to six months. The allow you to do that better? Then look at what
more complex solutions will often begin with a minimum you can afford for what you’re trying to achieve.”
viable product (MVP), the first and smallest deliverable that
can be built that is meaningful to the user and tests the – Robert Gibby, Chief Talent Scientist, Talent Acquisition, IBM
assumptions made by the business. These can then be quickly
enhanced to bring even more value to the business. With the
Five steps to getting started
APIs that are available today, even complex projects can show
a return in less than one year. Importantly, achieving results in IBM’s experiences of AI deployment in HR enable us to draw
fast timeframes requires a mindset change. Design thinking insights that could support any organization on their own AI
and agile working approaches will help to rapidly build adoption journey. These can best be summarized as the five
prototypes and iterate towards improved versions in short steps to getting started, as shown in Figure 4.
periods of time.
Nature of AI project
If you are using AI to improve hiring, you might see some
benefits quickly, such as a faster time to hire. However, other 1. Start with a business case
anticipated benefits, such as higher performance, will take
longer to occur. A retention solution will also take time to 2. Decide to buy or build
show value, since you need to wait to see whether or not
3. Identify the skills you have and need
workers leave or don’t leave in the way you expect. There are
other areas where you can see results within as little as six 4. Implement MVP
months. Chatbots, for example, can be built and deployed very
quickly. IBM’s performance management chatbot was built 5. Roll out enterprisewide
and deployed in under four weeks.
Page 13
Step 1: Start with a business case Step 2: Decide whether to buy or build
Companies need to start with understanding the business One question that organizations need to think through is
problem they are trying to solve with AI. The problem should whether they will build their own AI applications or buy
be one that can be resolved with improved insight, information, off-the-shelf AI applications for HR. Unless your organization
and data. Ask who the people are in the business line, not in has deep experience in AI, it is often better to leverage the
HR, that need this problem solved and will commit to it in their efforts of companies that have already developed solutions.
profit and loss reporting. For example, hiring might be a big If you decide to build your own solution, a good way to start
issue for your management team because they know they are is to solicit ideas from the workforce. IBM took this approach
growing and need to continue to grow. If the business can’t in a Cognitive Build competition. Employees were asked to
afford to increase the HR staffing to meet this demand, you generate ideas for AI applications, and the best were selected
have an obvious business problem that everyone can rally (via employee crowd-sourcing) and funded. Watson Career
around with an AI application. Coach, described above, is an example of an AI application that
resulted from this process. An important contributing factor to
Creating a business case will help you think about the the success of this approach was bringing the users of the apps
structures you need in place, and ensure you are focused into the design process.
on the right problem. Once you have this focus, you can ask
what skills and investments you need to solve the problem, Importantly, regardless of whether you buy or build, deploying
and you can build a business case. Keep in mind that while your AI capability in the cloud is the best way forward. Moving
many organizations are used to thinking about HR solutions to the cloud forces you to think about the data you have and
in functional capabilities, HR professionals and leaders will the data you need, and how to close the gap between these
need to flip their thinking. Instead they should think about the two perspectives. Cloud HR systems are also important to
decisions that managers, employees, and candidates make in successful AI in HR because the most powerful cloud-based
their work routines and how they could make better decisions AI applications share data from systems that were
if they had better insights. This approach may require input traditionally distinct.
from across multiple HR sub-functions.
Page 14
Step 3: Identify the skills you have and need Irrespective of whether you buy or build, HR practitioners
The skills required for implementing AI in HR can be using AI need to be much more knowledgeable about the
categorized into skills required for development, and skills business. Where they have been functionally siloed in the
required for implementation. If you are working with a vendor past, they now need to look at the employees in their entirety,
rather than developing your own AI application, then only the because AI applications often work across functional areas
implementation skills will be necessary. (e.g., benefits can be traded for compensation, assignments
overseas can now integrate with career development).
Development: Consider first the development skills. To
develop AI for HR in your own organization you will usually Step 4: Implement the MVP
need a multidisciplinary team to come together. The key skills IBM sometimes refers to a ‘self-funding’ model. Self-funding
to have in that team include: someone with an understanding refers to the idea that the business cases have projected
of the market and the business problem; development and revenue increases or projected cost savings that cover the
engineering skills; a technical architect who understands AI, costs associated with the project. Once these objectives are
data, and which APIs to use; HR domain expertise; and finally, met, the AI project has then paid for itself. In other words,
a design lead who understands how to present technology the business makes an investment today for savings next year
from a user experience perspective. These are the core roles or the following year, so the first steps take some fortitude and
for designing and developing AI applications in HR, and they boldness. You wouldn’t proceed if the business case wasn’t
are the skills that IBM puts on any team developing an AI strong. This approach requires someone to invest the
HR application. first dollars at the outset, but this is done with the clear
expectation of a return outlined in the business case. This
Implementation: When it comes to implementation and approach also needs an HR leader with confidence that the
ensuring adoption, a different set of skills is required. This results will get delivered, and that person will need strong
tends to be a less technical skillset. AI expertise is not influencing skills to make the case to the business. As time
required, but an analytical approach is needed. Also needed passes, and if the savings or revenue increases are sustained,
are good hypothesis forming skills, and the experience to ask there is usually more funding available. IBM typically reinvests
questions such as where the data came from, who trained the this funding in further development of AI applications.
AI application, what their motives are, and whether it is
appropriate to use the data in the way intended. In other The faster you deploy your MVP, the sooner you can deliver the
words, the implementers, and ideally the end users, need to next improved iteration. Ideally all projects, irrespective
be educated consumers of the technology. If you are using of complexity, will be like this and show benefits within 6 to
AI to deliver HR outcomes to employees, you need to ask 12 months. These timelines assume you have resources,
these questions to know whether the system has been dedicated effort, and data available. If you struggle to have
properly trained. Those responsible for ensuring adoption any of these elements, your timelines could be pushed out
need to know more than just how to turn the AI application further. For example, while chatbots are easy to put in place,
on. Also recommended is technical curiosity, the desire to they are ineffective without the content behind the interface
understand how the technology works and how to use it on which to base answers to questions from employees.
effectively. HR practitioners with this curiosity can provide
ideas for the next generation of AI HR applications.
Page 15
Step 5: Roll out enterprise-wide
If the MVP roll out is successful, scaling to the enterprise is the
next step. To check whether the MVP showed enough success
to warrant further expansion, use the ROI methodology we
discussed in the previous section. Every time IBM wants to
initiate or take a project to the next stage of development, say
from proof of concept to scaling, a business case is required.
The business case has to have the right level of sponsorship
from someone responsible for delivering the project. Business
cases need to be tied to revenue increases or cost savings, not
soft outcomes.
Page 16
Tips for successful AI adoption in HR AI and wider societal considerations
IBM has learned many lessons in its deployment of AI, AI systems need to be developed in an ethically sound
and interviewees recounted many that could be helpful to manner if the promise of AI is to be fulfilled while mitigating
practitioners. Some of the major learnings are shared here. the potential risks. For this reason, IBM complements its work
on the technical development of AI with policy streams on AI
– Don’t wait until you have the perfect solution and ethics covering five areas; accountability, value alignment,
It is better to release a minimum viable product and explainability, fairness, and user data rights7. In the HR
position it as such with your employees. When IBM context, ensuring AI applications are developed in an ethical
released the performance management chatbot, people manner ensures our applications are consistent with broader
knew it was ready but not perfect, and were encouraged to societal values and goals. We discuss some of these societal
ask questions that would challenge the bot. This allowed implications in this section.
IBM to deploy the AI solution quickly, while at the same
time enabling the chatbot to improve. Net effect of AI on jobs
Experts make all sorts of predictions about the net effect of AI
– Empower people with AI on overall job numbers. It is important to be transparent about
Designers of AI systems have a stewardship responsibility the fact that nobody knows for certain the end effect of AI on
to ensure that AI empowers workers. Designers should the number of jobs. It is impossible to make a prediction that
keep in mind that people feel most empowered when their is likely to be accurate. What everyone can be confident of,
decision autonomy is augmented rather than replaced. AI however, is that some roles will disappear, some roles will be
has been viewed most favorably when managers have the created, and almost all roles are going to change as a result of
option to override the AI recommendations when not seen AI. It is not the first time that technology has displaced jobs.
as optimal. IBM encourages managers to be comfortable Looking back, it would be unimaginable to argue, for example,
overriding the AI recommendations as appropriate. that agrarian technology should not have been developed in
order to protect the roles of farm workers. The same is true
– Ensure transparency
today with the use of AI in HR. In the short-term we don’t see
For managers to feel comfortable working with an AI
a lot of HR jobs going away that have not already disappeared
recommendation, it is important that there is clarity and
due to technology prior to AI (e.g., the internet). The jobs at
transparency about why AI recommendations are made.
most risk are very transactional jobs, such as people who ask
This should include making clear to managers and
pre-screening questions in recruitment.
employees what the recommendation aimed to achieve,
what data were used to make the recommendation, which Are chatbots taking jobs?
variables influenced the recommendation most, as well as Consider the AI applications that people might think are most
identifying all the variables on which the recommendations likely to replace jobs, chatbots. What IBM is actually seeing is
were based, and the expected accuracy of the that the chatbots are addressing basic questions, things that
recommendation. are routine, allowing HR professionals to address more complex
HR questions. In fact, jobs have been created in the process of
– Consider language and culture
deploying chatbots. These jobs relate to designing and training
AI requires context, reflected in data from different regions
the bot. For most jobs, responsibilities are changing rather than
where AI will be deployed, to learn and make appropriate
disappearing. It is components of jobs that are going away. The
recommendations. An AI solution designed in one region
parts that are going away are the rote, routine parts that can be
of a multinational organization’s operations may need to
automated. The parts of jobs that are gaining in importance are
be entirely retrained before deployment in another
analytical thinking, strategizing, and driving change. The key
region—even if the same language is spoken in both areas.
questions are how the way work gets done is going to change,
This relates to both the development and training of the
which roles will experience a lot of change, which will
model, as well as the experiences of the end users of the
experience little change, and how quickly all of that is going to
AI solution.
happen—short or medium term? Answers to these questions
– Design each app with other apps in mind can help provide the information that workers will need to
As you develop your AI solutions, it is a good idea to have re-skill, in order to establish career paths in the more
a holistic view of the end goal in mind. This will avoid, for compelling roles that are emerging in HR.
example, having a proliferation of unconnected chatbots all
addressing related questions but not drawing on common
infrastructure and data. It is increasingly important to do
this because apps often do not sit in the traditional siloed
HR subdomains. For instance, the compensation example
discussed earlier requires information from compensation
and learning, among other HR functions.
Page 17
AI creates higher value jobs Use of historical information
There are a number of higher value jobs that are being A critical element in IBM’s approach to increasing diversity
created in HR due to AI. The new Talent Scientist role to date has been to be very careful with how it uses historical
integrates sourcing, industrial-organizational psychology information concerning characteristics of top employees. These
and AI to analyze data, identify top talent, leverage enhanced top performers have a long list of characteristics, but only some
decision support, and predict success—all in a way that of these will be related to how well they perform. If you focus
appeals to candidates. In other words, they marry the on high performer characteristics unrelated to performance,
science of data to the art of sourcing, and continually drive a not only will you be making poorer decisions, but the AI will
successful hiring outcome with higher predictability. Another have the perverse effect of harming diversity. This is because
new role at IBM known as Talent Influencers makes up the the bias is learned, reapplied, amplified, and made systemic.
next generation of digitally and socially connected recruiters For example, if your top performers are all male, and you
who have a personal brand tied to domain and industry. They select more males because of this, you are likely to have a
use cutting-edge platforms popular with candidates to attract discriminatory hiring system. You are focusing on an attribute
top talent. of your high performers—gender—but it is not an attribute that
relates to performance. There is no theoretical reason that
IBM’s HR departments are now staffed with data scientists, gender should be a predictor of job performance, and this view
and the company has a whole team focused on sentiment is supported by analyses at IBM.
analysis and listening to internal social media for chatter to
gauge business sentiment. This would have been unheard The people writing apps at IBM work hard to ensure that the
of before AI was implemented in HR. Designers and user applications are feature blind to aspects that are not related to
experience experts are becoming more prevalent in HR future job performance. AI can also help remove bias through
as well. Learning scientists with AI skills work alongside analysis of decisions of managers. IBM can show where there is
neuroscientists to reimagine IBM’s learning content and a deviation from expectations and show statistically what the
experience. In short, HR can now offer more compelling causes of the disparities were.
careers as the prevalence of the human-machine partnership
continues to increase. Build in fairness and ensure transparency
It is unlikely that AI will resolve all issues related to bias.
AI technology simply augments a person’s capabilities, and
“I believe AI is giving a boost to different kinds the capabilities can be used for good or bad. This needs to
of jobs, and much more compelling jobs and be reflected in the way we think about AI applications, from
the very foundation of their design. AI solutions need to and
careers, including within the HR function.”
can be fair and unbiased by design. Building fairness into your
– Obed Louissaint, Vice President Human Resources AI systems, and being completely transparent in your use of
for Watson, Cognitive Solutions & Research, IBM AI (whose data, what the information is used for, etc.) will
ensure you stay on the right side of the good versus bad
technology divide.
AI, diversity, and bias
AI has great potential to help with bias if we are careful
about design. Companies developing these systems and
the organizations deploying them have a responsibility to be
conscious of potential bias, and to work actively to address
its existence. There is nothing about AI that magically reduces
biases. AI technology can be used to reduce bias or, if we
aren’t careful, reinforce it. At IBM, there are many areas
where AI shows promise at reducing bias. For example,
candidate attraction and recruitment AI solutions, described
earlier, create a mindset shift in candidates about what’s
possible by recommending jobs they did not know they were
qualified for. This is because these applications focus on skills
in the recommendation and hiring system. If you have the
skills required for a role, these applications tell you about the
role, irrespective of your gender or ethnicity. Building apps
in this way has required close working between content
domain experts and machine learning teams—both skill sets are
required for effective AI. Diversity of thought in the creation
of AI apps is essential to building apps that are as free of bias
as possible.
Page 18
A multi-pronged approach to increasing diversity IBM Smarter Workforce Institute
and inclusion at IBM
The IBM Smarter Workforce Institute produces rigorous,
Attracting and hiring diverse candidates to IBM is a top priority
global, innovative research spanning a wide range of workforce
and requires a mix of technology and training. In addition to
topics. The Institute’s team of experienced researchers applies
ensuring that AI tools are not reinforcing or introducing bias,
depth and breadth of content and analytical expertise to
IBM has infused AI into the hiring process to ensure talent
generate reports, whitepapers, and insights that advance the
pipelines remain diverse and inclusive. This includes having
collective understanding of work and organizations. This paper
women and diverse ethnicities play an equal part in the
is part of IBM’s ongoing commitment to provide highly credible,
creation of the algorithms.
leading edge research findings that help organizations realize
– Attraction: At the start of the hiring process, IBM uses value through their people.
home-grown AI and machine learning tools to help craft
job descriptions free from gender bias. How IBM can help
– Sourcing: In the sourcing phase, IBM AI tools proactively IBM is a cognitive solutions and cloud platform company
source and find applicants from talent pipelines matching that leverages the power of innovation, data, and expertise
key success profiles, surfacing candidates who may have to improve business and society. By bringing together
been missed by recruiters, thus eliminating any steps in the behavioral science, artificial intelligence, and expert
search process that may have introduced unconscious bias. consulting, IBM helps companies attract, hire, and develop
– Screening: When screening candidates, IBM uses an the talent they need to grow their business. For more
inclusive set of AI algorithms to ensure recruiters are information, visit ibm.com/talent-management.
unbiased in recommending jobs by forcing group
characteristics like gender, race, ethnicity, age, and others,
to be invisible during screening.
– Interviewing: During a live interview between a hiring
manager and a candidate, IBM allows AI tools to ‘listen in’
for indicators that help predict a candidate’s performance.
The AI tool then prompts the interviewer with question
suggestions that reduce the chance for unconscious bias
during the interview.
Page 19
Acknowledgments About the authors
The authors would like to thank the following Nigel Guenole, Ph.D. is an Executive Consultant with the
for their valued contributions to this report: IBM Smarter Workforce Institute and a Senior Lecturer in
Management at Goldsmiths, University of London. He is known
Obed Louissaint for his work in workforce analytics, statistical modeling and
Carrie Altieri
Vice President, HR, IBM psychological measurement. Nigel’s work has appeared in
Vice President, HR
leading scientific journals including Industrial Organizational
Communications, IBM
Tina Marron-Partridge Psychology: Perspectives on Science and Practice and
Elyse Anchel Global Leader, Talent & Frontiers in Quantitative Psychology & Measurement, as well
Director, HR Business Transformation, IBM as in the popular press. Nigel is also co-author of the book
Development, IBM The Power of People: Learn How Successful Organizations
Richard McColl Use Workforce Analytics To Improve Business Performance
Debora Bubb Global Leader, Talent (Pearson, 2017).
Vice President, Learning Technology, IBM
& Inclusion Officer, IBM Sheri Feinzig, Ph.D. is the Director, IBM Talent Management
Nicholas Santaniello Consulting and Smarter Workforce Institute and has over 20
Joanna Daly Digital Marketing years’ experience in human resources research, organizational
Vice President, HR, IBM Strategist, IBM change management and business transformation. Sheri has
applied her analytical and methodological expertise to many
Gordon Fuller Anshul Sheopuri
research-based projects on topics such as employee retention,
Vice President, Chief Vice President, HR, IBM
employee experience and engagement, job design and
Learning Officer, IBM organizational culture. Sheri has presented on numerous
Tom Stachura
occasions at national and international conferences and has
Diane Gherson Vice President, HR, IBM
co-authored a number of manuscripts, publications and
Chief Human Resources
Drew Valentine technical reports. Sheri is also co-author of the book The
Officer, IBM
Vice President, People Power of People: Learn How Successful Organizations Use
Robert Gibby & Culture, IBM Workforce Analytics To Improve Business Performance
Chief Talent Scientist, (Pearson, 2017).
IBM Amy Wright
Managing Partner,
Kelli Jordan Talent & Transformation,
Director, Career & Skills, IBM
IBM
Haiyan Zhang
Nickle LaMoreaux Lead Researcher, IBM
Vice President, HR, IBM
Jon Lester
Director, Transformation
& Automation, IBM
Page 20
© Copyright IBM Corporation 2021
IBM Corporation
New Orchard Road
Armonk, NY 10504
Produced in the United States of America
October 2021
IBM, the IBM logo, ibm.com, and Kenexa are trademarks of International
Business Machines Corp., registered in many jurisdictions worldwide. Other
product and service names might be trademarks of IBM or other companies.
A current list of IBM trademarks is available on the Web at “Copyright and
trademark information” at www.ibm.com/legal/copytrade.shtml.
This document is current as of the initial date of publication and may be changed
by IBM at any time. Not all offerings are available in every country in which IBM
operates.
THE INFORMATION IN THIS DOCUMENT IS PROVIDED “AS IS” WITHOUT ANY
WARRANTY, EXPRESS OR IMPLIED, INCLUDING WITHOUT ANY WARRANTIES
OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND ANY
WARRANTY OR CONDITION OF NON-INFRINGEMENT. IBM products are
warranted according to the terms and conditions of the agreements under which
they are provided.
1. Combs, J., Liu, Y., Hall, A., & Ketchen, D. (2006). How much do high-
performance work practices matter? A meta-analysis of their effects on
organizational performance. Personnel Psychology, 59(3), 501-528.
2. Ployhart, R. E., & Hale Jr, D. (2014). The fascinating psychological
microfoundations of strategy and competitive advantage. Annual Review of
Organizational Psychology and Organizational Behavior, 1(1), 145-172.
3. Cappelli, P., & Keller, J. R. (2014). Talent management: Conceptual
approaches and practical challenges. Annual Review of Organizational
Psychology and Organizational Behavior, 1(1), 305-331.
4. Lewis, R. E., & Heckman, R. J. (2006). Talent management: A critical review.
Human Resource Management Review, 16(2), 139-154.
5. Guenole, N., Ferrar, J., & Feinzig, S.L. (2017). The Power of People: Learn
How Successful Organizations Use Workforce Analytics to Improve Business
Performance. Pearson FT Press.
6. Guenole, N. & Feinzig, S.L. (2017). Decoding Workforce Analytics: A simple
guide to research design and analysis. IBM Smarter Workforce Institute.
Accessed at: http://ibm.biz/decode_HRanalytics.
7. IBM (2018). Trusted AI for Business. Accessed at:
https://www.ibm.com/watson/ai-ethic