Artificial Intelligence

and Improving the

Customer Experience

Vince Jeffs,
A PEGA
Director Strategy & Product Marketing,
Build
WHITEPAPER
Pegasystems
for
Change
Contents
4 The AI Imperative: Practical Application
6 Think Big, Start Small
8 Focus on Results First
11 Why Your Data Matters
12 Defining the Rules of the Game
14 Choose Your Vendors Wisely
17 Final Thoughts

2 WWW.PEGA.COM
Introduction
The term “artificial intelligence” (AI) is seeing No matter how you define AI, it is here
a dramatic resurgence, leading to both to stay, and will only become more
potential confusion and powerful ways to pervasive. From smarter phones to
use it for customer engagement. In fact, AI smarter dwellings to plotting our daily
has been around for many years, in many actions, AI offers the collective promise
forms. Yet only in the past decade has and reality of automation and intelligence
the door truly opened up to its practical from technology. The potential impact on
application – a change fueled by the new our personal and work lives is massive.
abundance of data to power decisions According to IDC (and depicted in Figure 1),
as well as ever-increasing customer “Widespread adoption of cognitive systems
expectations. and artificial intelligence (AI) across a broad
Brands that are in the business of range of industries will drive worldwide
improving customer experience for their revenues from nearly $8.0 billion in 2016
clients now need to cut through the hype to more than $47 billion in 2020.”i
and hyperbole. Organizations must figure The market for cognitive /
out what matters and where they can
AI solutions will experience a
safely place their bets for real results.
compound annual growth rate
(CAGR) of 55.1% over the 2016-
2020 forecast period.i

Figure 1:

$47
billion

55% CAGR Automated & Intelligent

(Cognitive) Software
Applications & Platforms

$8
billion

2016 2020
What’s clear is the opportunities are immense, if only businesses can harness their power.

3 WWW.PEGA.COM
The AI Imperative: Practical Application
While it’s important for brands to The evidence suggests that many
understand what AI is in order to have organizations have yet to deliver on these
some sense of its purpose and potential, simple AI use cases. A study by Bain & Co
it’s more important to have a clear found that 80% of CEOs believed they were
understanding of how your business wants delivering a superior customer experience,
to engage its customers. This requires yet only 8% of their customers agreed.ii
an honest appraisal of your current Moreover, when marketing executives
engagement model and a pragmatic view were asked in a recent survey how well
into where it must move both near and their organizations fulfilled promises they
long-term. made to customers, only 16% said they
For example, are marketing treatments performed optimally.
highly personalized to the individual, or Assuming your business may be among the
simply blasted out to customer segments? other 84%, you need to first envision how
Are salespeople empowered with insights to transform your customer experience,
about their accounts and contacts that declare that as your destination, and put a
they wouldn’t otherwise know, or are they reasonable time horizon on getting there.
forced to enter data into a tracking system Indeed, most businesses today plan very
that gives them nothing in return? Are few aspects for their business beyond
customer support agents equipped with three years, as market and customer
guiding intelligence to help them delight dynamics alike simply move too fast. This
customers during every interaction, or are is why businesses are now increasingly
they bombarded with data that they need adopting practical AI blueprints that they
to interpret on their own? can realize in two to three years, and start
getting value from in just a few months.

4 WWW.PEGA.COM
To achieve this, you must first envision Use cases for delivering value along a
a set of smarter, more proactive and customer’s journey abound. However, real
personalized engagements with your experience-wide adoption only comes
customers. These new interactions when consumers themselves are open to
must be achievable and made to deliver these techniques. Recent research shares
measurable value quickly. For example, a that there is good news in this respect:
bank may wish to anticipate banking needs
In a recent Pega survey of 6000
their customers will have within the next
adults, only 28% said they were
six months and then proactively present
options to them, even before the customer uncomfortable with a business
may be aware of their needs. The bank using AI to improve interactions.iii
might use AI to achieve this vision in the Businesses that sense these
following ways: experience gaps and present relevant
ƒƒ A parent may not realize that now is recommendations related to those needs
the time to start planning and saving will get the consumer’s attention. Those
for their child’s college education and that just spew random offers, proving they
their own retirement, but using AI, the know nothing about their customers, will
bank can sense that need and help that quickly turn them away.
individual. AI technology using smart data sources can
ƒƒ Some customers unfortunately don’t help you understand where each customer
pay their bills, putting them at risk for is in their unique journey, and then tailor
defaulting or even going bankrupt. The to their situation. Lay out these use cases,
bank can use the power of AI to predict and bake these into your AI plans.
which customers are likely to become
delinquent, and preemptively execute
tactics that steer them toward payment
and solvency.
ƒƒ A baby boomer with a complicated
financial situation may not have
executed five critical documents for an
estate plan, but the bank can detect
this speed bump in the customer’s
journey and engage them with these
services, adding real value and utility.

5 WWW.PEGA.COM
Think Big, Start Small
You may be tempted to use AI to solve
dozens or even hundreds of problems
right away. However, the businesses that
drive the fastest and best results focus
on tackling the issues that will have the
greatest impact on customer experience.
They strategically select forms of AI to For the fastest and
make their products and services smarter,
faster, and easier to buy and use. best results, focus on
To follow this best practice, identify the
low-hanging fruit – the higher-impact,
tackling the issues that
lower-effort, less invasive opportunities to
transform your brand for optimal customer
will have the greatest
engagement. For example, you could:
impact on customer
ƒƒ Reduce mundane customer service
tasks using automation technologies, experience.
like chatbots or robotic automation, to
free up customers from the shackles
of wait times, which at the same time
frees your client-facing employees to
handle higher-order work.
ƒƒ Predict which accounts and contacts
are most likely to influence and buy,
allowing salespeople to focus on the
right opportunities.
n
Risk

ƒƒ Present relevant promotions to

customers based on their individual
needs rather than the needs of the
business, driving higher conversion
rates and better customer satisfaction. Reward

6 WWW.PEGA.COM
Remember: AI isn’t some magical
technology that’s coming in a future
release. It’s here now. Practical application
of real AI today pays huge dividends.
Economic impact studies commissioned by
major technology providers demonstrate AI isn’t some magical
that it’s not only possible but also highly
likely, you could attain: technology that’s
≈≈ 300% to 600% increase in coming... it’s here now.
customer response rates
≈≈ 10% to 50% reduction in
customer churn
≈≈ Net Promoter Score (NPS)
increases from 10 to 40 points
≈≈ 300% to 500% Return on
Investment (ROI)
In fact, a recent Forrester TEI studyiv
showed that a company using Pega’s
AI-infused customer engagement
applications achieved:

Figure 2:
Financial Summary Showing Three-Year Risk-Adjusted Results

Improved
ROI: NPV: Payback:
retention
438% $ 138 million 4.7 months
$139 million

Source: Forrester Research, Inc.

Sprint, one of the USA’s largest mobile

carriers, was able to achieve a 10%
reduction in mobile subscriber churn using
Pega’s real-time AI. That system was up and
running in just 13 weeks. Using an agile
approach and a quick win mentality, results
can happen in months, not years.

7 WWW.PEGA.COM
Focus on Results First
For buyers, AI has become an increasingly that humans can’t do, can’t do well, or
confusing technical landscape, with terms can’t do very fast. The upper portion (the
such as machine learning (ML), chat Automation layer) contains functions that
bots, virtual assistants, robotics, natural mainly mechanize things. The lower layer is
language processing (NLP), deep learning where the true advanced intelligence rests,
(DNN/RNN), supervised learning, image in the forms of predictions and forecasts –
recognition, text analytics, reinforced higher-level cognition.
learning, complex event processing (CEP), To meet your business goals, you need to
and more seemingly emerging every day. see the big picture. Instead of spending
All you need to know is that each of energy on mapping capabilities, work
these technologies is a building block for backward from outcomes. For example,
solving meaningful use cases, and each your business might require:
ultimately helps you accomplish some ƒƒ More loyal customers.
task. Figure 3 provides an illustration of
ƒƒ Who will promote their great
these building blocks and their relationship
experience to others.
to specific outcomes. These approaches
are grouped to make sense of the many ƒƒ And who will continue to buy from you
AI technologies that help businesses because you offer great products and
automate burdensome and repetitive tasks are easy to do business with.

Figure 3:

8 WWW.PEGA.COM
Yet chances are high you aren’t firing on Figure 4, read from right to left, shows how to
all of those cylinders. To start, identify your work backward from your desired outcome.
biggest challenge, and then work backward Since the issue is customer satisfaction, your
from the areas you need to address to focus should be upon:
improve the customer experience. For ƒƒ Improving the web / mobile app
example, assume your greatest problem experience.
isn’t acquiring customers, but keeping
ƒƒ Improving the support experience.
current customers happy.
ƒƒ Making sure each customer is getting a
commensurate value for the product or
service they purchase.
Figure 4:

Machine learning (ML): Computer programs that can improve Supervised learning: A type of machine learning where a teacher
performance (learn) when given feedback without being explicitly provides the model with a training data set, which it uses to formulate
reprogramed. its algorithm, and then once the model is put into practice, it refines its
prediction or classification via additional supervision.
Chat bots: An interactive computer-messaging program that conducts
a conversation via an auditory or textual interface, designed to simulate Image recognition: The ability of software to identify objects, places,
how a human would behave as a conversational partner, able to people, writing, actions, and so forth, in a digital image or video.
exchange pleasantries, answer questions, and offer other means of
Text analytics - The ability of a computer program to classify, interpret,
assistance.
translate, and glean insights from structured or unstructured text data
Virtual assistants: A software agent that can perform tasks or services from any source.
for an individual, such as sending an email or scheduling an appointment.
Reinforced learning: A type of machine learning where an operator
Robotic process automation (RPA): Software that allows a company provides the model with a performance goal and ongoing feedback via
to record repeatable processes in an existing application, that otherwise rewards as to whether there is progress toward goal achievement, but
would require human input, such as via a keyboard or mouse, and then no other explicit instructions on the behavior expected to meet the goal.
trigger automatic execution when required, By doing so, these programs As such, the model experiments with behaviors, and learns which are
can automatically process a transaction, input and manipulate data, superior based on the reward system.
trigger responses, and interface with other applications.
Complex event processing (CEP): Software that takes multiple inputs
Natural language processing (NLP): The ability of a computer program (signals) and tracks whether an activity pattern has been satisfied or
to record, classify, interpret, translate, and glean insights from human not, based on an accounting of signals and pattern formulation. For
speech as it is spoken. example, a credit card firm may use a CEP system to detect fraudulent
behavior by monitoring what it deems to be unusual patterns of credit
Deep learning: A branch of machine learning that uses layers of neural
card use behavior.
networks (mimicking the design of the human brain) where outputs of one
layer become the inputs of the next layer. These neural models have proven
very good at nebulous classification tasks such as image recognition.

9 WWW.PEGA.COM
Work back from
business outcomes
to identify your
greatest challenge,
and opportunity, for
positively impacting
customer experience.

Pega designed this framework as a you improve support and raise customer
template that your business can leverage satisfaction while simultaneously lowering
for your AI aspirations. As you go through operating costs.
this exercise, ask yourself how your Your bot would then go through various
business could infuse more proactive stages of evolution. At first, its main job
intelligence into any crucial area. may be to keep valuable agent time
In fact, by using AI you may get a “double from being wasted by answering simple
positive” effect. For instance, if you decide questions. Eventually, its functional
that improving customer satisfaction is intelligence may increase until it’s capable
your biggest priority, then one important of taking some rather nebulous issue
step could be using AI to introduce an statements and effectively routing to the
effective chatbot experience (because you right people who can most efficiently
know your customers have asked for 24/7 resolve the problem.
support with limited friction). As you do so,

Figure 5:

10 WWW.PEGA.COM
Why Your Data Matters
Anyone who’s taken a computer- As you navigate your customer experience
programming course has heard the phrase, (CX) focal path, take stock of the data each
“garbage in, garbage out.” AI systems are area should provide, and then crosscheck
no different. They use data as fuel – and with the data you have. Run simple tests
if it’s incomplete, tainted, inaccurate, or to gauge the effort to source the data,
biased, then low octane levels will lead to a feed it into an AI CX pilot, and keep a
sluggish, underperforming AI. running tab of the issues you encounter.
However, many things have changed since This list becomes your inventory of data
that programming idiom arose. Systems projects you’ll need to complete to ensure
source massive datasets in various your AI CX effort won’t choke due to lack
formats at blazing speed; cleansing of proper fuel.
routines can root out errors, fill missing You also need to pressure test whether
fields, cross-correlate sources, and these existing data management
append new, related data. systems can hold up under the new load
Of course, high quantities of unstructured, of AI. Also, don’t forget that for most
raw data can sometimes lead to slower organizations, small data – the CRM
processing. As a result, organizations information you have on your customers,
still find themselves wrangling massive your transaction systems, and your
amounts of raw data and struggling to product catalog – remains an untapped
make it usable and accessible. resource. You can start improving
customer retention using machine
In order to implement an effective learning algorithms with small data (e.g.,
customer engagement solution, AI a 50-field data set). Don’t get hung up on
must have memory banks full of solving your big data problems before you
great information. Designers have the start tackling real business solutions.
opportunity to wire in information coming
from raw data pertaining to customer
behavior, profiles, device type and usage,
and much more.

In order to implement an effective customer

engagement solution, an AI must have
memory banks full of great information.

11 WWW.PEGA.COM
Defining the Rules of the Game
AI as an entity evolves only when it can Ultimately, AI is a tool for humans, which
sense, remember, take some action, and is why it has two popular modes of
learn from the results, in a continuous operation:
cycle as shown in Figure 6. In order for it to (1) In some cases, AI can operate
be a fully functioning and orderly system, somewhat independently, with its
it must include a decision-making layer, outcomes monitored by its creators. Its
which contains institutional memory and decisions are often superior in terms of
processing steps. In fact, according to a scale, speed, and balance to the decisions
Forbes article citing Forrester research, its makers would have made.
the single most valuable use of AI for
organizations today is in the area of (2) In other cases, AI can provide adjunct
decision management.v intelligence to humans, guiding and
assisting them, improving eventual action
That decision-making sub-system has paths and outcomes. One common
critical functions and guardrails – it example is providing frontline employees
serves as “the rules of the game.” For with real-time next-best-action guidance,
businesses, that can include a host where staff members leverage AI to predict
of deterministic constraints, such as the best thing to say next during a live
never taking certain actions outside interaction. These real-time intelligent
of regulatory guidelines, limiting systems are primed with customer
promotions to available capacities to behaviors, risk and value scores, and
fulfill, or offering products that fit the predictions on response likelihood, and
risk profile of the client. While machine when woken are fed contextual data,
learning is powerful, AI is only effective arbitrate between varieties of possible
for enterprises if business leaders can actions – even factoring in economic value
effectively monitor and fine-tune these – and return an optimal next step.
rules to optimize AI in the right direction.

Figure 6:

Predict &
DECIDE

LEARN Customer-Centric ACT

SENSE

12 WWW.PEGA.COM
These recommended actions deliver products and services to match their needs,
indisputable results: Brands have routinely provide educational materials, or resolve
seen two to six times improvements in their problems. One of the world’s largest
their response rates. Take, for instance, telecommunication providers, Vodafone,
Sprint, whose CEO Marcelo Claure publicly won a prestigious Stevie award for their
announced, “We’re seeing more than “Pusula” applicationvii (“Compass” in English),
double the acceptance of retention offers. which guides more than 1300 associates
That is a big deal.”vi toward delivering better offers and better
Similarly, store employees working side customer experience. The result, displayed
by side with customers also benefit in Figure 7, was year on year revenue growth
from AI engines. Equipped with tablets, of over 60 percent, and increased NPS (Net
they can help customers find the right Promotor Score) by over 20 points:

Figure 7:
Continuous revenue growth and stronger customer experience with Pusula

13 WWW.PEGA.COM
Choose Your Vendors Wisely
Should you trust one vendor to lead
you to the promised land? The short
answer is probably not, especially if you
are a very large enterprise, because
one single solution doesn’t exist. In a
recent webinarviii, Olive Huang, Research
Director for Gartner, referred to the
inevitable “System of Systems” that every
large organization will build as they piece
together a better solution for customer
engagement.
Anyone who followed the venture capital
spree that resulted in an explosion of AI
and customer experience startups knows
that regardless of how you try to organize
it, there’s a complex set of players – each
with strengths and weaknesses.

The AI and customer experience landscape is

home to a complex set of players – each with
some strengths and weaknesses.
Chances are good that your company
is already doing business with over 100
vendors claiming some AI capability for
better customer experience. That’s too
many to manage effectively, especially when
many have overlapping capabilities, and
only adds confusion that keeps you from
achieving your customer experience goals.
For this reason, brands should strive to
rationalize their unique vendor landscape,
and double down on several stable, agile,
innovative vendors with chops for the short
and long haul – ones that play unique and
pivotal roles.

14 WWW.PEGA.COM
When selecting a vendor, keep these ƒƒ Ensure a centralized AI
recommendations in mind: decision authority:
ƒƒ Seek quick wins while being While most businesses will eventually
cautious of big bang approaches: employ AI from a variety of different
Your efforts to infuse AI into all aspects vendors, success will hinge upon
of your customer’s journey can and leveraging ONE centralized customer
should start quickly, with a focus on a decision hub – a single brain, serving
defined interaction. This lets you return as the central source for intelligent
value in a continuous, connected, and decision-making and arbitration across
agile fashion. Look for a vendor that all channels.
can deliver rapid value in short spurts ƒƒ Demand proof of scale:
(90-120 days) and will still be a trusted If you are a large business with millions
partner in the years ahead. They of customer conversations happening
should be able to provide a blueprint every day, you need a vendor that has
of how these wins fit together and will proven their software can scale to these
lead to long-term transformation. levels and perform well. If your vendor
ƒƒ Demand proof of AI outcomes: can’t provide proof of scale, chances
Many vendors are now clamoring are very good you will be an extension
for attention using AI as a buzzword. of their performance-testing lab – and
Unfortunately, not all have a proven you’ll pay them to test their software.
track record of the customer
engagement outcomes a business
may require. Some may even be just a
marketing veneer covering a stack of
acquired and disparate software and
platforms. Require extensive proof
points and customer evidence of AI
driving the specific outcomes your
business is seeking.

15 WWW.PEGA.COM
Final Thoughts
AI has enormous potential to affect our We can marvel at the potential of AI, but
lives in positive ways. For consumers, it can ultimately we should act now to reap its
automate monotonous tasks and assist in practical rewards. Predictive intelligence
difficult ones. For businesses, it can provide technologies can play a critical role in
a competitive edge, differentiate products digital transformation – helping your
and services, and even contribute to business garner customer loyalty, improve
success or failure. experiences, and ultimately increase profits.
As humans on both sides of this commercial
marketplace, we are in control – meaning
we are responsible for the actions of our
machines. We must feed them, program
them, and monitor them wisely, or face
hidden and overt consequences.

i
IDC, Worldwide Semiannual Cognitive/Artificial Intelligence Systems Spending Guide, 2016
ii
CMO Council, Predicting Routes to Revenue, 2016
iii
Pegasystems, Artificial Intelligence Survey, 2016
iv
Forrester, Total Economic Impact (TEI) of Pega Marketing, https://www.pega.com/insights/
resources/forrester-total-economic-impact-tei-pega-marketing, 2016
v
Forbes, https://www.forbes.com/sites/gilpress/2017/01/23/top-10-hot-artificial-
intelligence-ai-technologies/#4b6d209d1928, 2017
vi
Fortune, http://fortune.com/2015/08/07/sprint-pegasystems-customer-service/, 2015
vii
Stevie Awards, http://stevieawards.com/sales/vodafone-turkeys-pusula, 2016
viii
Gartner, Customer Engagement Hub Part 1 – What is it and why do you need one?, 2016

16 WWW.PEGA.COM
ABOUT PEGASYSTEMS
We are Pegasystems, the leader in software for customer engagement and operational excellence. Our
adaptive, cloud-architected software – built on the unified Pega® Platform – empowers people to rapidly
deploy, and easily extend and change applications to meet strategic business needs. Over our 30-year
history, we’ve delivered award-winning capabilities in CRM and BPM, powered by advanced artificial
intelligence and robotic automation, to help the world’s leading brands achieve breakthrough results.

For more information, please visit us at WWW.PEGA.COM

© 2017 Pegasystems, Inc. All rights reserved. All trademarks are the property of their respective owners.