Get Started With Artificial Intelligence

Fast-track your AI explorations by learning

from early adopters.

Info-Tech Research Group Inc. is a global leader in providing IT research and

advice. Info-Tech’s products and services combine actionable insight and
relevant advice with ready-to-use tools and templates that cover the full
spectrum of IT concerns.
© 1997-2021 Info-Tech Research Group Inc.
Table of
Contents
3 Executive Brief Give Your AI a Purpose

5 Executive Summary 91 Phase 3: Select Your First AI Proof of

Concept
32 Phase 1: Explore the Possibilities
109 Summary of Accomplishment
33 Step 1.1: Understand AI Trends and
Technologies 112 Bibliography

50 Step 1.2: Understand Organizational

Impact of AI

55 Phase 2: Learn From Your Peers and

Info-Tech Research Group | 2

Get Started With Artificial Intelligence
Fast-track your AI explorations by learning
from early adopters.

EXECUTIVE BRIEF
Analyst
Perspective We are at the dawn of a new era in digital transformation. The fourth
industrial revolution is about aligning AI with core areas of the
business. The transformation requires embracing cultural and
Welcome to the fourth industrial organizational shifts.
revolution AI is rapidly growing with machines intelligently executing high
performance and dynamic tasks such as driving cars and detecting
fraud. Companies have much to gain by harnessing AI’s strategic
advantage. Seventy percent of companies are expected to have
adopted AI in their business within this decade (McKinsey Global Institute,
2018).

Ambitious start-ups use the power of AI technology to drive increased

efficiency, engage customers, and obtain automated data-driven
decisions. Can you afford to ignore this revolutionary force that is
shaping our world today?
Anuradha Ganesh
Principal Research Director - Data, AI and
ML
Info-Tech Research Group
Executive summary
Artificial Intelligence Challenges Common Obstacles Info-Tech’s Approach
• Understanding what AI really means in practice • Established rigid business structure • Use this blueprint to understand what AI really
means in practice and to get started with AI to
• Identifying specific applications of AI for your • Fear of unknowns arising from AI
harness its transformative power.
business  technicalities
• Follow Info-Tech’s methodology and step-by-
• Understanding the type of AI technology • Lack of awareness and understanding of
step approach to give your AI a purpose and
applicable for your situation automation benefits
select your first AI PoC:
• Optimizing investments with highly dynamic • Lack of AI training opportunities in business
and growing AI technology
• Restructuring, re- and up-skilling workforce 1. Explore the Possibilities
• Applying AI for your competitive advantage
• Investment constraints for long-terms benefits 2. Learn from peers and give your AI a
Organizations across the globe are looking for purpose 
guidance on how to separate the signal from the Acknowledgment and reduction of existing 3. Select your first AI PoC 
noise when wading through the masses of obstacles are needed to take advantage of AI-
marketing material around AI. induced long-term gains for future growth.

Info-Tech Insight
Data-driven AI enables discovery of hidden structures and facilitates a dynamic data-adaptive ecosystem to unleash
opportunities for growth and innovation. Info-Tech Research Group | 5
Organizations are climbing the AI maturity ladder
Don’t get left behind.
Across all industries and in many departments, investments are being made in AI technology. Real benefits are being seen,
giving organizations advantages through greater efficiency, increased revenue, and enhanced customer intimacy. Where is your
organization on the maturity scale?
AI adoption and understanding, based on survey responses
18%

100% Pioneers
30%
90% 33% Investigators
34%
80% Experimenters
70% 21% Passives
60%
50% 16%
40%
47%
30%
Source: MIT Sloan Management Review, 2018
20% 20%
10%
0%
2017 2018
"If you are customer-facing, AI technology is no
Embedded at least one AI capability
Embeded in several parts of the organization
longer a differentiating factor, it’s a necessity.".
Piloting ‒ Zahra Zahid, Slalom Consulting
Source: “AI Adoption Advances Survey,” McKinsey & Company, 2018 Info-Tech Research Group | 6
Despite the growth in supporting AI technologies, organizations are still
struggling to adopt them for their use

Real AI Struggles "Next to fostering an understanding of AI

throughout the entire organization, it is
Despite the many promises of AI, like with any other new technology,
organizations are struggling to fully realize AI’s potential. The reasons important to get the top management’s attention
boil down to a lack of understanding of when these technologies and make them aware of and excited about the
should and shouldn’t be used, as well as a fear of the unknown. There potential of this set of technologies."
are reports of AI replacing many jobs, and these fears are rational as we
‒ Harald Rudolph, Head of Daimler Strategy, Daimler
have yet to see how this technology will impact the current job In MIT Sloan Management Review, 2018
landscape; however, fears like this accompany all revolutionary
technologies. A better understanding of how to work best with the
changes will help all of us through this revolution. 0% 10% 20% 30% 40% 50% 60%
Employee fear of change 54%

A large proportion of the challenges that Lack of in-house skills 54%

organizations face in adopting AI Lack of knowledge about where AI
49%
can assist
technology is in the understanding of
what AI is and how it will help at all Cultural acceptance 47%
levels – senior executives and Senior management resistance 37%
employees alike.
Source: Infosys, 2017

Info-Tech Research Group | 7

The world is changing quickly – but it is not too late to become a
part of the AI revolution
Don’t wait for the revolution to come to you.
The AI revolution is just getting started. There are limitless
opportunities for AI to improve how organizations work and to create
competitive advantage:
• 75% of organizations using AI technology report sales of new
products and services being boosted by more than 10%.
• 78% of organizations using AI technology report increasing
operational efficiency by more than 10%.
• 75% of organizations using AI technology report increasing
customer satisfaction by 10%. Source: Capgemini Consulting, 2017

AI helps organizations keep up with the (Dow) Joneses

Relative to competitors, respondents say their company’s adoption of AI has allowed them to:

16% 20% 27% 28% 9%

Catch up Stay on par Edge slightly ahead Widen the lead Leapfrog ahead

Source: Deloitte Insights, 2018 Info-Tech Research Group | 8

Executive
Summary its AI journey.
Info-Tech’s Know your enemy:
Approach and The first step in taking advantage of the Define your problem:
revolutionary new AI technologies out there
methodology is to know when AI is applicable. Although
After helping you understand what AI is and
the types of problems it can solve, this
most marketing hype would have you believe
blueprint will help you define the specific
that it is something that will solve all your
problem that your organization needs to solve
problems, AI technology is something that
and brainstorm how – and if – AI can help.
can help you solve a specific problem. AI
technologies are unsuitable in many Choose your AI path:
circumstances.
Based on your use cases, start small with a
Learn from others: PoC and prove your AI’s value as it grows.
Follow a use case‒based approach and the
lessons of others to determine the direction
that you should take your organization in on

Info-Tech Research Group | 9

 Insight 1 AI is not a magic bullet.
Insight Instead, it is a tool for speeding up data-driven decision making. A more appropriate description
of current AI technology is data-enabled, automated, adaptive decision support. Use when

Breakdown appropriate.
Insight 2 Garbage in, garbage out still applies to AI – and is even more
relevant!
AI technology has its foundations in data. Lots and lots of it. Rich, relevant, accurate, and
timely data from interconnected systems is essential to effective use of AI. How good is your
data? Are you ready for AI?
Data-driven AI enables
discovery of hidden Insight 3 Learn from your peers and early adopters.
structures and facilitates AI is a rapidly evolving field, and there are many published use cases with documented
a dynamic data-adaptive business value. Using a use case-based approach, you can learn from the successes and
failures of others to more rapidly narrow down how AI can deliver value for your
ecosystem to unleash organization.
opportunities for growth Insight 4 Start small to gain experience and to lay a solid foundation.
and innovation. Too many big projects have failed in the past. Start small with AI and then wash, rinse,
repeat. Keep in mind, though, that the real value will come from the enterprise adoption of
AI, and for that you will need to create an AI strategy.

Insight 5 Give your AI a purpose.

Your AI strategy and current exploration activities should closely align with the strategic
goals and drivers of your organization. The key question you should be asking is not
“What can AI technologies do?” but rather “What can they do for us?” and “How much
would we benefit from AI if we were to invest in it?” Info-Tech Research Group | 10
The hype: AI is everywhere

“AI: The Killer App for Your Business”

(Forbes. “Killer App,” 2018)

“Meet Your New Employee”

(HBR, 2015)

“The AI Doctor Will See You Now”

(WSJ, n.d.)

“Anything You Can Do, AI Can Do Better”

(The Economist, n.d.)

“Machines Will Replace Humans at the Top - And Wipe

Us Out if We Dare to Resist”
(Express, 2019)

“Intelligent Machines Will Replace Teachers Within 10

Years”
(Independent, 2017)

Info-Tech Research Group | 11

What’s in a revolution?
Revolutions are caused by technological advances that fundamentally reshape key aspects of the world, such as commerce, health care,
learning, and the environment. The following are some of the revolutions that have shaped who we are today and how our society
functions:

1700s-1940s 1940s-Present Present

First Industrial Revolution Scientific Revolution AI Revolution
Radio, aviation, and nuclear We are currently undergoing a
The mechanized spinning of
fission sparked the radical shift in how we engage
textiles, large-scale manufacturing
Scientific/Technical Revolution with each other, organize our
of chemicals, steam power, and
(1940-1970). society, and ultimately how we
efficiencies in iron-making sparked
know ourselves. Thanks to the
the first Industrial Revolution
Information Revolution building blocks laid down by
(1760-1840).
previous revolutions (electricity,
Second Industrial Revolution The internet and digital hyper-connectivity, and
media and devices caused/is communication technology, for
Railroads, the telegraph and
causing the Information example) and rapid improvements
telephone, electricity, and other
Revolution (1985-present). in AI technology, we can predict,
utilities sparked the second
automate, and interact in ways
Industrial Revolution (1870-1940).
that couldn’t have been imagined
in other revolutions’ times.
Info-Tech Research Group | 12
AI is transforming the world
Simulation of human intelligence by machines.
Artificial
Intelligence Mimics human cognitive functions such as
(AI) sensory perception, learning, and problem
solving.
Source: Wikipedia, “Artificial Intelligence,” n.d.

9.5 trillion – 15.4 trillion potential total annual value of AI

and analytics across industries.
Source: McKinsey & Company, “The Executives AI Playbook,” 2018

Three ways AI is transforming the world:

"Economic history has taught us that
the impact of major innovations is
often felt in the most unexpected
places." 
‒ Ajay Agrawal et al., Prediction Machines,
Intelligent automation 2018
Labor Business model disruption
augmentation Info-Tech Research Group | 13
AI is automating (away)…
The situation before AI Enter AI
In organizations in all of human history, humans have With modern AI technologies, business processes
made decisions. In modern organizations, this required are not only replaced, but also redefined. Taking in
defined processes for the intake of information and large amounts of data, algorithms can be created
business rules for taking certain actions in defined that make those business decisions based on
situations. This in turn made making decisions more identified patterns. Not only that, but the
efficient and provided opportunities for making smarter algorithms can adapt to incoming data to make the
processes. smartest decisions possible.
Occupations
So, what’s changing? • Radiologists
Processes
• Taxi and truck drivers
• Finance: AR, AP, expenses, reconciliation, and closure of financial ledgers; audit and
• Insurance claim adjusters
compliance; fraud detection; touchless transactions, such as automatic requisitioning of
• Cashiers
services/supplies from a third-party preferred supplier; provisioning of costs within the budget;
new project planning. • Librarians
• HR: Talent sourcing, selection, and recruitment. • Teachers
• Operations: Intelligent routing and processing of customer requests; workforce scheduling; • Bank tellers
predictive maintenance; inventory management; demand forecasting. • Executive assistants
• Legal: Legal discovery; contract reviews. • Translators/interpreters
• IT: Service desk, automated operations, security. • Reporters
• Product design: Generative design, content creation. • Postal workers
Info-Tech Research Group | 14
We are living a third wave of business
"~50% of current work activities are technically
automatable by adapting currently demonstrated
technologies." Now: Adaptation
‒ McKinsey & Company, “The Promise and Challenge of the Age of Artificial Intelligence,” 2018
Now the third wave is taking
place and it involves adaptive
1970-1990s: Automation
processes. It is ushering new,
innovative ways of doing
The second wave consisted of business. Advances in AI open a
Early 1900s: Standardization automated processes. It started in world of possibilities to reimagine
the 1970s and peaked in the current processes to be more
1990s, with the business process flexible, faster, and adaptable to
The first wave took place during re-engineering movement, thanks the behaviors, preferences, and
the early 1900s, and it involved to advances in information needs of the workers at a given
standardized processes, technology: desktop computers, moment.
culminating in assembly line and large databases, and software that
overall processes that could be automated various back-office
measured, optimized, and tasks.
standardized to achieve
efficiencies.
Source: Daugherty and Wilson, 2018

Info-Tech Research Group | 15

AI is augmenting humans
Physically: collaborative robots
Rather than being confined to dedicated enclosed areas as
in the past, robots on the assembly line these days work
side by side with humans, being aware of where the
humans are and what they are doing. They typically still
lift and place heavy components while humans execute
tasks that require dexterity and judgement, but they are
also able to comprehend and learn thanks to machine
learning.

Image source: Tech Advantages. Used with permission

Cognitively:
"AI makes predictions cheap, and
predictions are at the heart of decision
making."
‒ Ajay Agrawal et al.,
Prediction Machines, 2018

Info-Tech Research Group | 16

“We are at the dawn of the new era in decision
making”*
* 
Predictions are at the heart of decision making and some predictions (and decisions) are already being fully automated, especially routine ones, where rich
data sets are available to train the algorithms. Machines are also better suited than humans to make predictions involving many input dimensions or a
complex interaction among the variables. Humans, on the other hand, are exceptionally good at making predictions when data is not available or limited,
and in situations requiring analogy, causal inference, or judgement.
When humans and machines are combined, the prediction accuracy typically goes up and surpasses that of just humans or just machines. In the future,
many decisions and actions will be completely automated. Others will be handled by the machine + human teams, finally delivering on the promise of data-
driven decision making.
Info-Tech Insight
Source: Eckerson Group, Anatomy of a decision-making process
“The Impact of AI on When decisions are based on
Analytics,” 2018 
probabilities, a mental shift is
required: from deterministic and
AI will transform BI and the way binary thinking (yes/no) to
people make decisions and act. Rather probabilistic and less certain. For
than starting with a hypothesis, data example, what does it mean that a
analysts will begin with an AI-driven likelihood of something is 95% or
insight. Instead of querying data to 82% or 70.4%? How would that
prove or disprove their hypothesis, translate into action? Which actions
users will query data to expand or would those be? Do decision makers
validate machine-generated insight or understand this? Are they ready to
recommendation – or they may act on operate with this new thought model?
the AI-based insight at face value.
‒ Eckerson Group, “AI: The New BI,” 2018
*Heading quote from Eckerson Group, Based on: Ajay Agrawal et al., Prediction Machines, 2018 Info-Tech Research Group | 17
“The Impact of AI on Analytics,” 2018 
AI is disrupting business
Many of the organizations that are household names today did not exist ten years ago. They took advantage of emerging
technologies to disrupt an established industry business model and/or create their own markets. And they continue to
innovate today with AI and machine learning.

Newcomers such Stitch Fix, a personalized shopping

service, continue to disrupt their industries with AI.
Stitch Fix has 80 data scientists and almost 100
engineers on staff headed by the Chief Algorithms
Officer (ex-Netflix) who reports directly into the CEO.
All of Stitch Fix’s business processes – from item
recommendation to resource and inventory
management and intelligent logistics – are driven by
data and algorithms. They even do “algorithmic
fashion design,” meaning that they identify gaps in Source: Stitch Fix, n.d.
their inventory of products and create designs to fill
those gaps. Info-Tech Research Group | 18
AI is gaining most traction in the areas of business that create the most
value within a given industry

Info-Tech Research Group | 19

Exhibit from “Notes From the AI Frontier: AI Adoption Advances, But Foundational Barriers Remain,” Nov. 2018, McKinsey & Company.
Why now?
We have all heard grandiose promises of technologies before. More often than not we are disappointed by the limitations or
reality of the technology. AI has been guilty of this. However, this time it’s different. Three realities in the business world
support this view:
Cheap AI really took off when researchers started to employ graphics
Cloud computing and commodity hardware have changed computing
the game when it comes to heavy number crunching. These power & processing units (GPUs) traditionally used in gaming. Google
are prerequisites for AI technologies. powerful recently developed a tensor processing unit (TPU)
hardware specifically for neural net machine learning. Intel and
The major cloud platforms offer pay-as-you-use-it
computing options, which means that you can scale up or Facebook are working on a chip to make it cheaper to use AI.
down as needed. Source: Medium, “Cheaper AI for Everyone,” 2019
1

Learning is the key to modern AI. It is

incredibly difficult to write a program that
can act on a strict logical basis with the vast AI sits on a foundation of large
amount of data that is generated every day. Real Artificial
Intelligence amounts of data, which is required
Instead, AI needs to be able to make for learning. Organizations today are
decisions about what information is capable of harvesting large volumes
2 3
important, and what is not. The current and of data of increasing varieties, and
constantly growing number of open-source Rapidly AI is how these organizations are
packages and libraries that allow people to improving Big data
capitalizing on their big data
leverage complex AI algorithms are a boon algorithms
Source: EVRY, 2017 investments.
to data scientists implementing AI
technology.
Info-Tech Research Group | 20
Applied AI is what the current state of AI is: the use of AI concepts or
techniques for a specific purpose
The diagram below demonstrates the two broad categories of AI, and how they relate to each other:
General Artificial Intelligence Applied Artificial Intelligence (AAI)
General AI (aka Good Old-Fashioned AI or GOFAI) is what most people Applied AI (AAI or Narrow AI) is the type of AI that is
think of when they hear about AI. This is the science fiction view of AI currently taking the world by storm. This involves the use of AI
where the intelligent being demonstrates a flexible understanding of its technologies, many of which are based on the fundamental
environment and has reasoning skills that allow it to traverse between traits of General AI, to solve a specific problem. AAI is all
vastly unrelated topics and use cases, just as a human can. This is what about predictions, and is nothing more than mathematics.
we want and envision AI to be. It’s a dream. Most experts agree that AI
hasn't yet achieved even a six-year-old human’s level of intelligence, let The four types of AAI are listed below:
alone that of a mature human. The traits of General AI are listed below:

Deduction and Machine

Reasoning
Machine Learning Computer
Learning
Vision
Knowledge
Perception (ML)
Representation
Natural Language Natural
Planning
Processing
Robotic
Language Process
Motion and
Social Intelligence
Manipulation
Processing Automation
(NLP) (RPA)
Source: Knowledge Representation Framework Project, 2010

"Any sufficiently advanced technology is indistinguishable from magic."

Info-Tech Research Group | 21
‒ Arthur C. Clarke (quoted in Medium, “The Future Is Indistinguishable From Magic,” 2018) 
AAI technologies are helping organizations make faster decisions and
predict future outcomes
We have
In this all heard we
blueprint, grandiose promises
will focus onofAAI
technologies
and its before, and more
practical often than not we are disappointed by the limitations or reality of
uses.
the technology. AI is an old culprit of this. However, this time it’s different. Three realities in the business world support this view:
The types of AAI and the algorithms underlying these technologies overlap to varying degrees and have common analytical
goals that support the goals of the business. All of the technologies can be applied individually or in combination with each
other. Applied Artificial Intelligence (AAI)
Analytical Goals Business Goals
Applied Artificial
Discover Automation
Intelligence (AAI) Machine
technologies are Learning (ML) Integration
helping Classify
organizations make Scale
faster decisions and
predict future
AAI Predict
Intelligence/
Insight
outcomes
Robotic Process Impact/
Automate
Automation Engagement
Computer (RPA)
Vision Natural
Language
Processing Info-Tech Insight
(NLP)
The merging of machine learning with any
combination of these three technologies
represents the sweet spot of AAI. Info-Tech Research Group | 22
Follow Info-Tech’s methodology to secure your place in the
future by leveraging AI technology
We have all heard grandiose promises of technologies before, and more often than not we are disappointed by the limitations or reality of
the technology.
When AI is anwith
getting started old culprit of this.
emerging However, thissuch
technologies timeas
it’s different. Three realities in the business world support this view:
Info-Tech will help you learn the fundamentals of AI
AI, it is important to get up to speed with the technological while targeting useful applications of AI for your
knowledge while also keeping the broader organizational organization:
context in mind. To implement effective AI, the organization and its members,
Broad business knowledge including senior members, must have a good understanding of
the principles of AI technologies, including:
Deep AI knowledge

• Machine learning (ML)

• Natural language processing (NLP)
• Computer vision
Result: • RPA
Effective AI used in the right context But most importantly, they need to understand how these
technologies could be used by their employees, customers, and
To implement useful AI, you have to think about the technology
partners; where they fit in the existing business processes (and
in the context of: what changes would be required); and what impact they would
• People • Data have.
• Processes • Customers Info-Tech Insight
• Technology architecture • Business partners AI technology without context is not
useful. Info-Tech Research Group | 23
Info-Tech walks you through the process so you will understand how AI can help
you and how to get fast value out of your AI
To get the most value out of AI, organizations must understand which AI technologies should be used in various situations
(and when they shouldn’t be used). They must create a prioritized set of projects based on business needs, starting small with a
PoC, then developing plans to scale up across the company.

This blueprint walks you through the following

What you will learn in each step:
framework to get the most value out of your AI
initiatives:
The current landscape of AI technology and how
1 organizations are using it to solve business problems.
AI Technology Globally Use this knowledge to do the most important step in an
AI project: ask the right questions and define your AI’s
purpose.
What others in your industry are doing with AI to
2
Industry Implications revolutionize how they automate, predict, and engage their
customers and clients.

Organizational Plans What to do in your organization to create the most value from
your AI initiatives based on the global and industry implications
above.
3
Info-Tech Research Group | 24
Airbus used AI to ramp up
production of a new jet
INDUSTRY SOURCE
Aeronautics Sloan Management Review, “Reshaping
Business With Artificial Intelligence,”

Case Study 2017

Airbus A350 Reduction in time spent

33% dealing with manufacturing
When Airbus introduced its new A350 jet, the company knew that it disruption.
would need to bring in the big guns to ramp up production.
Disruptions in production of a brand-new jet are commonplace, and
for a multi-billion-dollar project, disruptions needed to be kept to a
minimum.

AI to the Rescue

To minimize production disruptions, the aircraft manufacturer used

previous manufacturing data combined with a learning algorithm to
identify and predict patterns in manufacturing disruptions.
Results
Using this technology, the manufacturer was able to predict 70% of
issues in real time. As a result, the time it took Airbus to deal with
disruptions was shortened by one-third.

Info-Tech Research Group | 25

Key deliverable: Blueprint deliverables
Each step of this blueprint is accompanied by supporting deliverables to help
Get Started with Artificial you accomplish your goals:
Intelligence Deck

Contextualize, identify the purpose of your

Understand AI Trends and
AI, and create a minimum viable business
Technologies Idea Reservoir Tool
case (MVBC) for your first AI PoC for
submission to the executive sponsor and
the governing bodies. 

• Analyst will help to fine-tune the value

Prototyping
proposition and will share best-practices to
maximize the likelihood of approval. workbook
Idea
I List:
I I I I I
d d d d d d
e e e e e e
a a a a a
a
Contextualize and develop 1 2 3 4 5
6
your AI PoC Project Ideas
Minimum Viable Business
8 2 5 4 3 1 Case (MVBC) document
0

Info-Tech Research Group | 26

Info-Tech offers various levels of
support to best suit your needs

Guided Implementation
DIY Toolkit Workshop Consulting
“Our team has already made this “Our team knows that we need to “We need to hit the ground “Our team does not have the time
critical project a priority, and we fix a process, but we need running and get this project or the knowledge to take this
have the time and capability, but assistance to determine where to kicked off immediately. Our project on. We need assistance
some guidance along the way focus. Some check-ins along the team has the ability to take this through the entirety of this
would be helpful.”  way would help keep us on over once we get a framework project.”
track.” and strategy in place.”

Diagnostics and consistent frameworks are used throughout all four options.

Info-Tech Research Group | 27

Info-Tech’s methodology for Get Started With Artificial
Intelligence – Project
2. Learn From Your Peers 3. Select Your First AI
1. Explore the Possibilities
and Give Your AI a Purpose PoC
3.1 Develop Your AI PoC Project
2.1 Understand How Your Peers Are Using
1.1 Understand AI Trends and Technologies Ideas
AI
Phase Steps 1.2 Understand Organizational Impact of AI 3.2 Create an MVBC for Your First AI
2.2 Give Your AI a Purpose
PoC

• Ideas developed, socialized, and

• Relevant use cases identified and
Good understanding of benefits, prioritized.
contextualized.
Phase Outcomes opportunities, and limitations of AAI. • First PoC selected, MVBC
• Project team defined.
created and submitted for
• Mandate and charter created.
approval.

Info-Tech Research Group | 28

Guided implementation A Guided
What does a typical GI on this topic look like? Implementation (GI) is a
series
of calls with an Info-
Phase 1
Phase 2
Phase 3
Tech analyst to help
Explore the Possibilities
Learn from your Peers and give
your AI a Purpose
Select your First AI PoC implement our best
practices in your
organization.
Call #1: Call #3: Select AI Call #5: Identify Call #6:
Discuss AI use cases that may resources to be Discuss
trends and apply to your involved in the project ideas. A typical GI is between
technologies organization and
with the discuss their
project and
Call #8: eight to 12 calls over the
document the
analyst and get potential value. mandate and
Discuss course of four to six
MVBC for
your questions Discuss challenges charter. Call #7: first PoC. months.
answered. with implementing
Document
these AI use cases
Call #2: Identify dependencies
in your
what AI means to and
organization.
the organization Call #4: Discuss the determine
and the drivers. strategic plans of your feasibility.
organization and establish a Select first
business context for AI. PoC.
Info-Tech
Info-Tech Research
Research Group| 29
Group | 29
Discuss organizational
 Workshop Overview
Contact your account representative for more information.
[email protected] 1-888-670-8889

Pre-workshop Day 1 Day 2 Day 3 Day 4

Prepare Educate Establish AI Use Cases and Identify AI PoC and Draft a Create a Roadmap for
Potential Strategies MVBC Executing AI PoC

0.1 Execute Data Culture 1.1 Review Info-Tech’s Data 2.1 Discuss how AI is 3.1 Brainstorm your AI 4.1 Create AI mandate and
Diagnostic. Maturation Journey and impacting internal corporate project(s). charter.
0.2 Review current analytics path from BI to AI. functions: Accounting and 3.2 Create a PoC shortlist and 4.2 Discuss AI governance and
strategy. 1.2 Explore AI: review Finance, Operations, HR, select your first AI PoC. integration with data and
Activities

0.3 Review organization’s technologies and their use IT, and Legal. 3.3 Identify PoC prerequisites, analytics governance.
business and IT strategy. cases (UCs). 2.2 Discuss UCs you have dependencies, and 4.3 Create and prioritize key
1.3 Review AI UCs applicable identified as applicable to determine feasibility. initiatives for AI PoC.
0.4 Review other supporting
documentation. to your industry. your organization. 3.4 Review and complete Bias 4.4 Confirm resources to be
0.5 Confirm participant list for 1.4 Review your current 2.3 Align your AI initiative to Risk Assessment for involved in your AI PoC.
workshop. business plan/strategic the organization’s strategy selected project.
objectives and brainstorm and business plan.
where AI would fit to 2.4 Review primer on human
deliver value. and machine biases.
1.5 Identify UCs that may
apply to your organization.

1. Data Culture Diagnostic 1. List of applicable UCs and 1. AI alignment to 1. MVBC for AI 1. AI initiative charter
report potential PoCs organizational goals 2. Bias risk assessment for AI 2. Governance considerations
Deliverables

PoC for AI
3. Team RACI
4. High-level roadmap for PoC
execution

Info-Tech Research Group | 30

Suggested Reading

1. Daugherty, Paul R., and H. James Wilson. Human + Machine: Reimagining Work in the Age of AI. Harvard Business Review
Press, 2018.
2. McAffee, Andrew, and Eric Brynjolfsson. Machine, Platform, Crowd: Harnessing Our Digital Future. WW Norton, 2017.
3. Broussard, Meredith. Artificial Unintelligence: How Computers Misunderstand the World. The MIT Press, 2018.
4. Davenport, Thomas H. The AI Advantage: How to Put the Artificial Intelligence Revolution to Work. The MIT Press, 2019.
5. Agrawal, Ajay, et al. Prediction Machines: The Simple Economics of Artificial Intelligence. Harvard Business Review Press,
2018.

Product images sourced from Amazon

Info-Tech Research Group | 31

 This phase will walk you through

Phase 1
the following activities:

1.1 Understand AI Trends and

Technologies
Explore the Possibilities
• Understand the difference
between general AI and AAI.
• Understand technologies
involved.
• Learn how and where these
technologies are used.

Phase 1 Phase 2 Phase 3 1.2 Understand Organizational

1.1 Understand AI Trends 2.1 Understand How Your 3.1 Develop Your AI PoC Impact of AI
and Technologies Peers Are Using AI Project Ideas • Use cases on how AI
1.2 Understand 2.2: Give Your AI a Purpose 3.2 Create an MVBC for technologies are being leveraged
Organizational Impact of AI Your First AI PoC to automate, augment, and
transform business functions
within Accounting, Finance, HR,
Operations, Corporate Legal, and
IT.

This phase involves the following

participants:

Get Started With Artificial Intelligence • CIO, CTO, CDO, and CAO
• IT executives and managers
looking to learn about AI
Info-Tech Research Group | 32
• Analytics practitioners
1.1 Understand AI Trends and Technologies
Estimated Time: 2-3 hours

1. Start with reviewing the materials in this section: 4. Discuss what AI means for your organization and the business
1) Review important AI concepts and terms. drivers for AI.
2) Understand what AI is in the context of the modern
organization.
3) Understand the difference between general AI and AAI. Outcomes of this step:
4) Understand the technologies involved.
5) Learn how and where these technologies are used.  A solid understanding of what AI really is in the modern
world.

2. Discuss AI technologies and trends.  A clear picture of the types of problems that AI can help you
3. Get your questions answered and any outstanding issues Info-Tech Insight
solve.
clarified. AI that we read about is a dream.
AAI is what’s used in reality, and it is
highly contextual and narrow in
scope.

Info-Tech Research Group | 33

The current state of AI is applied AI: the use of AI concepts or techniques
for a specific purpose
The diagram below demonstrates the two broad categories of AI and how they relate to each other:
General Artificial Intelligence Applied Artificial Intelligence (AAI)
General AI (aka Good Old-Fashioned AI or GOFAI) is what most people Applied AI (AAI or Narrow AI) is the type of AI that is
think of when they hear about AI. This is the science fiction view of AI currently taking the world by storm. This involves the
where the intelligent being demonstrates a flexible understanding of its use of AI technologies, many of which are based on the
environment and has reasoning skills that allow it to traverse between fundamental traits of General AI, to solve a specific
vastly unrelated topics and use cases, just as a human can. This is what problem. AAI is all about predictions, and is nothing
we want and envision AI to be. It’s a dream. Most experts agree that AI more than mathematics.
hasn't yet achieved even a six-year-old human’s level of intelligence, let
alone that of a mature human. The traits of General AI are listed below: The four types of AAI are listed below:

Machine
Deduction and Computer
Reasoning
Machine Learning Learning
Vision
(ML)
Knowledge
Perception
Representation

Planning
Natural Language Natural Robotic
Processing Language Process
Social Intelligence
Motion and Processing Automation
Manipulation (NLP) (RPA)
Source: Knowledge Representation Framework Project, 2010

"Any sufficiently advanced technology is indistinguishable Info-Tech Research Group | 34

from magic."  ‒ Arthur C. Clarke (quoted in Medium, “The Future Is Indistinguishable From Magic,” 2018) 
Another point of view on AI

AI: Systems that extend human capability by sensing, comprehending, acting, and
We have all heard grandiose promises of technologies before, and more ofen than not we are disappointed by the limitations or reality of
learning.
the technology. AI is an old culprit of this. However, this time it’s different. Three realities in the business world support this view:
‒ Daugherty and Wilson, Human + Machine, 2018

What we want What we’ve got Actually, more like this

Image source: Library of Congress

Is AI like artificial flight?
Is AI a revolution in progress, or has it already
"The quest for ‘artificial flight’ succeeded when the
happened?
Wright brothers and others stopped imitating birds and
started … learning about aerodynamics,” Stuart Russell
Food for and Peter Norvig write in their leading
"People worry that computers will get too smart 
and take over the world, but the real problem is that they’re
thought textbook, Artificial Intelligence: A Modern Approach. too stupid and [computers have] already taken over
“AI started working when it ditched humans as a model the world."
[…]. Airplanes don’t flap their wings; why should
‒ Pedro Domingo, The Master Algorithm, 2015
computers think?" Info-Tech Research Group | 35
Interlude: A note on analytics, data science, and AI
AI and dataall
We have science are intimately
heard grandiose promiseslinked. When understanding
of technologies before, and more how tothannot
often leverage
weAIaretechnology
disappointedfor your
by the own organization,
limitations or reality of it
is imperative to involve
the technology. AI is anthe
olddata science
culprit of this.function ofthis
However, yourtmeorganization. Machine
it’s different. Three learning,
realities as used in
in the business the support
world businessthiscontext,
view: is a
form of (business) analytics. Just like data science is an evolution and extension of data analytics, machine learning extends
and enriches the data science (and hence the analytics) toolbox.
The Evolution of Analytics

Exhibit from Eckerson Group, “AI: The New BI,” 2018. Used with permission.
Info-Tech Research Group | 36
Interlude: So, what does a data scientist do?
The responsibilities of a data scientist include:
• Turning raw data into insight.
• Using statistics, machine learning, analytical, and visualization
approaches to solve business problems.
• Having strong programming skills, the ability to create new algorithms,
handling big data, and possessing expertise in one or several business
domains.
A data scientist should also be able to tell a story hidden in the data and
thus they must possess excellent communication skills.
A typical data scientist workflow is illustrated on the right. Note the
iterative nature of the Explore, Pre-Process, and Modeling steps. It may
take several trials to arrive at the right model. Data science is part
science, part art!
 Download the
Data Scientist job
description template

With AI and data science skills in short supply, it may be Exhibit from Wolfram Research, “Building a Data Scientist Pipeline,”
2017.
challenging to get started with AI. There are options, however, Used with permission.
from leveraging vendors to crowdsourcing; hosting a Kaggle
competition or a hackathon etc., while growing internal
competencies within your organization, which should be your
priority. No Data Scientist? – No Problem; You Ca
For more information and ideas, see Info-Tech’s note n Still Get Your Feet Wet With AI!
Info-Tech Research Group | 37
AAI technologies are helping organizations make faster decisions and
predict future outcomes
We have
In this all heard we
blueprint, grandiose promises
will focus onofAAI
technologies
and its before, and more
practical often than not we are disappointed by the limitations or reality of
uses.
the technology. AI is an old culprit of this. However, this time it’s different. Three realities in the business world support this view:
The types of AAI and the algorithms underlying these technologies overlap to varying degrees and have common analytical
goals that support the goals of the business. All of the technologies can be applied individually or in combination with each
other. Applied Artificial Intelligence (AAI)
Analytical Goals Business Goals
Applied Artificial
Discover Automation
Intelligence (AAI) Machine
technologies are Learning (ML) Integration
helping Classify
organizations make Scale
faster decisions and
predict future
AAI Predict
Intelligence/
Insight
outcomes
Robotic Process Impact/
Automate
Automation Engagement
Computer (RPA)
Vision Natural
Language
Processing Info-Tech Insight
(NLP)
The merging of machine learning with any
combination of these three technologies
represents the sweet spot of AAI. Info-Tech Research Group | 38
Machine learning is powering AI
Machine learning involves algorithms designed to give computer systems the ability to learn, or become better at their
tasks, through the use of data and without being explicitly programmed. The following represents some of the popular
algorithms* within the three types of machine learning:
• Labelled Data *Note: this list of algorithms is not exhaustive
• Direct Feedback
Supervised Learning • Predict Outcome/Future Reinforcement Learning
Algorithms Algorithms
• Q-Learning
• Decision trees
• SARSA (State-Action-Reward-
• Regression (linear, logistic)
Supervised State-Action)
• Support vector machines
• DQN (Deep Q Network)
• Naïve Bayes
• DDPG (Deep Deterministic
• K-Nearest Neighbor Policy Gradient)
• NAF, A3C, TRPO, PPO, etc.
Unsupervised Reinforcement
• No Labels
• No Feedback • Decision Process
• Find Hidden Structure • Reward System
• Learn Series of Actions
Unsupervised Learning Source: Tech Jini,
2017
Algorithms
• Clustering • Expectation maximization • Blind signal separation • Component analysis
• Anomaly • Hidden Markov models • Association rules Info-Tech Research Group | 39

detection
Supervised machine learning: learning from
examples
In supervised machine learning, humans teach the algorithm what conclusions it should come up with. It requires labeled data – for example, a
dataset containing customer attributes (e.g. age, gender, location, types of products they have, tenure) as well as labels indicating, say, who has
chosen to stay or leave a bank. The algorithm learns which features/attributes in the input data are associated with these labels.

Examples: Classifying labeled photos, filtering spam based on previously labeled spam emails, predicting whether an applicant will default on a
loan; predicting churn, etc.

Info-Tech Research Group | 40

Image: Exhibit from Hortonworks, “Data Science With Hadoop,” 2014. Used with permission.
Unsupervised machine learning: pattern detection

Unlike supervised learning, unsupervised machine learning does not require labeled data (hence “unsupervised”). In fact, in unsupervised learning, the
learning algorithm is never told what the data represents; its objective is to learn from the previous examples that are given. It does that by discovering
the underlying structure of the data and the patterns present there. In that, unsupervised learning is closer to what we expect AI to be like: be able to
learn/discern complex patterns without a human providing guidance.

Unsupervised algorithms are used when we don’t know what patterns the data contains, or what the outcomes should be. Unsupervised algorithms mine
input data for rules, detect patterns, and summarize and group the data points to derive meaningful insights and describe the data better to the users.
(There is also a category of semi-supervised learning, and many popular methods blend unsupervised and supervised learning.)

Examples: customer segmentation in

marketing, purchasing behavior, social
network analysis, etc.

Clustering methods group data

based on similarity so that data
points in the same group are more
like other data points in the same
group and dissimilar to the data
points in other groups. It is Association rule learning – Just like the name
Image: Exhibit from UX Collective, “An Intro to Machine
basically “birds of a feather flock Learning for Designers,” 2014. Used with permission.
suggests, these algorithms discover (”learn”)
together.” relationships (rules) hidden in large datasets, like
in the market basket analysis application.
Info-Tech Research Group | 41
Reinforcement learning: reward-based learning by trial and error
Reinforcement learning is like unsupervised learning in that no labeled input data is provided. However, like supervised learning, feedback is being given
– the algorithm (called the agent) is graded, just like a student, on performance. This feedback consists of rewards and punishments for the agent’s
behavior while it performs actions and interacts with the environment (typically games). The agent’s objective is to maximize the rewards and minimize
the risk. Learning happens by trial and error from the agent’s own actions and experiences and the feedback given. Reinforcement learning is typically
combined with deep learning/neural networks.

Examples: AlphaGo, traffic lights control, robotics (a robot picking up and moving an object), optimal trade execution in finance.

Reinforcement learning Agent: Mathematical model.

model and key Environment: Physical world in which the agent operates.
components: State: Current situation of the agent.
Reward: Feedback from the environment.

Exhibit source: Kdnuggets, 2018. Used with permission.

A reinforcement learning problem can be best explained through games. Let’s

take the game of Pac-Man where the goal of the agent (Pac-Man) is to eat the
food in the grid while avoiding the ghosts in its way. The grid world is the
interactive environment for the agent. Pac-Man receives a reward for eating
food and punishment if it gets killed by the ghost (loses the game). The states
are the location of Pac-Man in the grid and the total cumulative reward is Pac-
Man winning the game. Source: KDnuggets, 2018

"This is similar to how a dog learns to sit, or how

we encourage a child to play a piano through Exhibit source: Kdnuggets, 2018. Used with permission.
strategically given M&Ms." Info-Tech Research Group | 42
‒ Raia
Hadsell
Deep learning aka (deep) neural networks: learning by example
Deep learning is a separate class of machine learning algorithms. These algorithms can learn in a supervised and an unsupervised manner. Most modern
deep learning methods are based on artificial neural networks, which are inspired by the structure of the brain: biological neural networks. An artificial
neural network consists of two visible layers: a layer of input signal neurons where the information enters the neural network and a layer of output signal
neurons where the results are known; plus, several hidden layers in between, which transform the data. The word “deep” refers to these hidden layers,
which range from a handful to over a 100 layers (deep).
Deep learning also employs a technique called backpropagation (backprop) which adjusts the weights between the nodes (working backwards, hence
“backprop”) so that input leads to correct output.*
Biological Neuron Neural Network

Exhibit source: KDnuggets, 2017. Used with permission.

Exhibit source: Wikipedia. “Neuron3,” Public domain.

Deep learning algorithms have been around since the 1960s and really Some examples of where deep learning is used:
took off in 2012 when a team from the University of Toronto won the
Recommender Optical character
ImageNet competition using Convolutional Neural Networks. This Self-driving cars
systems recognition
Cancer detection
started the “deep learning revolution” that has transformed the AI
industry. Source: Wikipedia, “Deep Learning” Industrial Text and speech
Image recognition Drone surveillance
automation analytics
*For a very accessible introduction to
neural nets and backprop, see this article “Is AI Riding a One-Tric Info-Tech Research Group | 43
from MIT Technology Review. k Pony?”
Use natural language processing (NLP) to drive customer service and improved customer
engagement
NLP involves text or speech data, and it is behind popular AI technologies (such as chatbots) in driving improved customer
engagement.
Text data is unstructured, so it does not lend itself easily to traditional analyses associated with structured databases. Early NLP
approaches used symbolic methods and involved complex sets of handwritten rules. Machine learning methods were introduced in
the late 1980s and revolutionized NLP. Natural Language Generation
Natural
Language
Understanding
Phonology
Morpholog
y
Syntax

Semantics

Discourse
Pragmatic
s
Exhibit source: Automated Insights, 2016. Used with permission.

While chatbots might be the most popular NLP/NLU use case, other applications of these technologies include:
Customer sentiment analysis, spam filtering, topic identification and intelligent routing, text categorization, automatic
labeling of content, automatic content generation, text summarization, machine translation, info extraction (e.g. in drug Info-Tech Research Group | 44
Case Study
Improving customer satisfaction INDUSTRY
Transportation
SOURCE

Celaton, n.d.; Daugherty and

with machine learning and NLP Wilson, Human + Machine, 2018

The opportunity
At Virgin Trains, a train-operating company in the UK, the team of customer
service representatives used to manually read, sort, and route customer
complaints (emails). This process was slow, prone to errors, and highly labor
intensive. Customer satisfaction suffered as a result.

Enter AI – inSTREAM™ ML/NLP platform

The new platform reads and categorizes customer communications, extracts
sentiment and key data, and packages all this into a case-ready file for an
employee to review and process.

Results
The daily processing time and manual labor was reduced from 32 person-hours Reduction in manual effort
per day to four. The team can now handle surge periods without additional and daily processing time
85%
resources and process complaints from many channels including social media.
Customer satisfaction has improved..
Info-Tech Research Group | 45
 Use computer vision to get insights into the physical world
and automate processes
Image processing is driving increased insight and automation.
Computer vision is tightly linked with machine learning and focuses primarily on the unstructured information in image or video format. It seeks to automate
tasks that human vision can do, from classifying images, to identifying and localizing objects, analyzing motion, computing a 3D model of a scene, extracting
high-dimensional data to produce information to support decisions, etc.

Classification Instance
Classification + Localization Object Detection Segmentatio
n

Cat Cat Dog, Man, Cat Dog, Man, Cat

Image source: Catalysts. Used with

Single Object Multiple permission.
Objects
Examples of uses of computer vision with machine learning: Image source: Yonah. Used with permission.
• Automatic inspection in manufacturing applications.
Source: Tata Consultancy Services, 2017 • Human interaction – e.g. as the input to a device for human-computer
• Assisting humans in identification – e.g. a species identification system.1 interaction.3
• Controlling processes – e.g. in autonomous vehicles or with an industrial • Modeling objects or environments – e.g. medical image analysis or
robot. topographical modeling.
• Detecting events – e.g. for visual surveillance or people counting.2 • Navigation – e.g. by an autonomous vehicle or mobile robot.
• Facial recognition technology – e.g. Facebook’s automatic person • Organizing information – e.g. for indexing databases of imagesInfo-Tech
and image
Research Group | 46
tagging. Sources: 1Wikipedia, “Automated Species Identification”; 2Wikipedia, “Peoplesequences.
Counter”; Wikipedia, “Human-Computer
Automate the boring stuff with RPA Benefits of RPA-enhanced processes:
• Inexpensive/scalable/fast
RPA paired with machine learning speeds up heavily manual • Consistent and verifiable
activities. • More accurate
RPA software automates repetitive, rules-based processes usually performed by people sitting in • More fair, less subject to social
front of computers. By interacting with applications just as a human would, software robots can biases
open email attachments, complete e-forms, record and re-key data, and perform other tasks that
mimic human action.
While most RPA platforms do not yet include machine learning, ML + RPA is a natural and
powerful pairing. Intelligent process automation (IPA), as the field is known, “mimics activities
carried out by humans and, over time, learns to do them even better.” (McKinsey & Company, 2017)
RPA is deployed in: Machine Learning
Manufacturing
Data Entry and Make human-like
Supply Chain
Collection judgements based
Management
on historical data
Inventory and Stock
Invoice Management and training.
Management
Improved
Approximately 25% of processes are accuracy and
conducive to RPA enhancements. variety of data
that can be
Source: Tata Consultancy Services, 2017 automated. For more information on RPA, see Info-Tech’s
Info-Tech Insight blueprint Automate Work Faster and
RPA is typically the most feared type of AI due to the assumption that by automating tasks, computers will replace More Easily With Robotic
humans in their jobs. However, it must be stressed when implementing RPA technologies that RPA will lead to Process Automation
enhanced job creation (as seen with all disruptive technologies), as well as transformation of existing jobs to be
Info-Tech Research Group | 47
more value-adding.
Many technologies behind AAI are readily available for
use today
AI is a highly dynamic field, with many sub-disciplines which are evolving fast. Below is a list of just a few technologies that are
currently available and ripe for business adoption. The next phase of this blueprint lists many use cases across a range of industries to get
your creative juices flowing. As you consider where AI can deliver value to your organization, keep in mind that for AI to be successful,
it must be led by business, solve a problem that is well understood and self-contained, and finally, it needs to align with the strategic
objectives of your organization.

Available Today
Machine learning Emerging Technologies Future Technologies
(platforms)
Chatbots
Real-time emotion
NLP analytics

Image recognition Virtual companions "[…] history has

taught us that the
All-optimized hardware Thought-controlled impact of major
gaming
Digital twin innovations is often
Cognitive cybersecurity felt in the most
Biometrics unexpected places."
Neuromorphic computing
RPA ‒ Ajay Agrawal et al.,
Prediction Machines, 2018
Etc. Etc.
Info-Tech Research Group | 48
While AI may sound like magic, consider the
following:
Deep learning is, in essence, a pattern recognition and classification
engine.
Many recent successes in applying AI are due to advances in machine learning and
specifically deep learning/neural network algorithms. These algorithms are, however,
not intelligent, no matter how much you scale them up or how much data you use to
train them. On the contrary, they are greedy, brittle, opaque, and shallow (Source:
Wired, 2018):
• Greedy – They need huge quantities of data to train them. 
• Brittle – They work successfully only on very narrowly defined problems. For
example, an algorithm trained to recognize images of cats will not recognize other
animals and may even break if there are changes in the image background.
(Although a very dynamic field of transfer learning attempts to address exactly this
problem.)1
• Opaque ‒ Neural networks are black boxes, whose outputs are not easily explained, Deep learning’s advances are the product of pattern
since they consist of thousands of neurons arranged into hundreds of layers, where recognition: neural networks memorize classes of things
each neuron is carrying a weight that contributes to the final output of the
algorithm. (Explainable AI attempts to remedy this by lifting the lid off these black
and more-or-less reliably know when they encounter
boxes and explaining what’s happening on the inside.)2 them again. But almost all the interesting problems in
cognition aren’t classification problems at all.
• Shallow – They do not encode any of the knowledge we humans possess; they are
François Chollet, AI researcher at Google
essentiallySources:
mathematical/statistical
1 models.
Wikipedia, “Explainable AI”; 2
Wikipedia, “Explainable Artificial Intelligence”
(Source: Wired, 2018)

Info-Tech Research Group | 49

1.2 Understand Organizational Impact of AI
Estimated Time: 2-3 hours

1. Review use cases on how AI technologies are being

leveraged to automate, augment, and transform business
functions within Accounting, Finance, HR, Operations,
Info-Tech Insight
Corporate Legal, and IT.

Outcomes of this step: AI needs a purpose to be useful. Use

 A good understanding of how AI is impacting business this phase to learn the full potential of
AI technology today and start thinking
function.
what that the purpose is for your AI.
 Ideas about how your organization can leverage AI
technologies.

Info-Tech Research Group | 50

AI in Accounting, Finance, and
Operations "The back-office support system is being
dismantled to make way for the new
The use cases for AI in Accounting, Finance, and Operations, regardless of the industry, conception that is made possible by the
include the following: advent of artificial intelligence disruptive
• Automated data entry, categorization, and transaction processing. technologies which together promise speed
and significant cost reduction."
• Automated expenses.
‒ Shankar
Balasubramanian,
• Automatic reconciliation and closure of financial ledgers. HEXAWARE (Source:
EY, n.d.)
• Audit and compliance.

• Intelligent checking of invoices and automated payment of invoices to specified terms.

• Corporate budgeting, planning, and forecasting.

• Touchless transactions – e.g. automatic requisitioning (replacing bins, ordering new

school textbooks, etc.).
• Better management of the supplier base (search for best prices, best terms).

• Connected intelligence – e.g. contact center automatically requisitioning a third-party

preferred supplier to remove graffiti; provisioning for cost within the budget; automated
payment of invoices to specified terms.
• Optimal scheduling – e.g. routing of social workers or public works crews; optimal
Info-Tech Research Group | 51
maintenance.
Source: LGiU, 2018
AI in Corporate Legal: Freeing
people for higher value-add tasks

The use cases for AI in corporate legal departments,

regardless of the industry, include the following:
• Document classification for knowledge transfer,
retention and regulatory compliance
• Document review
• Contract review and management
• Legal research/discovery
• Due diligence

Mini use case: JPMorgan Chase & Co.

At JPMorgan Chase & Co., lawyers and loan officers used to spend hundreds of thousand of hours AI Meets Legal Research
every year studying commercial loan agreements in 12,000 new wholesale contracts – a mind-
numbing task. Now, an AI system called COIN (for Contract Intelligence) does the same job in
seconds and with fewer errors, freeing people for higher value-add tasks and allowing the capacity
to scale up and down as necessary. Info-Tech Research Group | 52
Source: Futurism, 2017
AI in HR: Hiring the right people fast, effectively, without bias
Recruitment/Talent sourcing challenges
An applicant’s skills, knowledge, and even experience are not good predictors of their future job success. References and unstructured interviews help
somewhat, but still not much. Also, men might apply for a job when they meet only 60% of the qualifications, while women might not apply for a job
unless they meet 100% of the qualifications. Soft skills such as communication, persuasiveness, adaptation, and conflict resolution are four times more
predictive (Source: Elevate 2018 Conference).
Men apply for a job Women apply for a job The solution? – AI!
when they meet only only if they meet
AI systems prowl LinkedIn, Glassdoor, Facebook, Twitter, etc. and proactively identify good
candidates. They help write job posts that are engaging and personalized using real-world
60% 100% hiring outcomes from millions of job posts and recruiting mails. And they help with screening.
Some draw on organization psychology and ask the applicant to fill out a detailed questionnaire
of the qualifications of the qualifications
to determine fit (using machine learning), others combine video intelligence.
Source: Elevate 2018 Conference.

Job seekers are filmed as they respond to Reduction in cost to fill vacancies
questions and are assessed for fit by analyzing 62%
their facial features and the answers. Reduction in time to fill
68%
Yet other applications can automatically More diverse candidates shortlisted
match an applicant with the right job for 23%
Source: Knockri, n.d.
them. (As many as 70% of workers may be in
positions for which they are not well suited.)
In addition to helping, employers find the best Managing workforce
candidate for a job, AI screening tools help to AI and machine learning systems are used to predict
reduce unconscious bias (such as the departures, identify internal candidates for open positions,
candidate’s sex, race, and/or age). assist succession planning, and support the HR help desk Exhibit source: Knockri, n.d. Used with permission.

(chatbots), etc. Info-Tech Research Group | 53

 AI in IT: From IT (Ops) to AIOps
Cybersecurity

DevOps
• ETL
• Enterprise
integration
• Alerting
• Software Help Desk
Image:
testing, etc. Chatbots
Darktrace, n.d. Production Operations
Used with
permission.

Info-Tech Research Group | 54

Exhibit: Logz.io, 2017. Used with permission.

 This phase will walk you through the
following activities:

Phase 2 2.1 Understand How Your Peers Are Using AI

• Understand how your peers are using AI.
• Identify use cases that may apply to your
Learn From Your Peers and Give Your AI a Purpose organization.
2.2: Give Your AI a Purpose
Get Started With Artificial Intelligence • Now that you understand what AI is, what
technologies are available for use, and how
your peers are deploying AI, this step will
help you determine the top three business
goals to get value from your AI.

This phase involves the following

Phase 1 Phase 2 Phase 3 participants:
1.1 Understand AI Trends 2.1 Understand How Your 3.1 Develop Your AI PoC • CIO, CTO, CDO, CAO
and Technologies Peers Are Using AI Project Ideas • IT executives and managers leading AI
1.2 Understand 2.2: Give Your AI a Purpose 3.2 Create an MVBC for efforts.
• Analytics practitioners.
Organizational Impact of AI Your First AI PoC
• Project/program manager(s)

Phase Outcomes
• An understanding of where AI is delivering value
in your industry and other verticals.
• Identification of applicable use cases.
• A list of business problems or opportunities that
AI could potentially help solve within the
Phase 2 Results & Insights:
• For AI to be of value, you need to give it a purpose. To ask the right questions of your AI, you need business understanding and context. organization.
• An understanding of when AI should be used and
• The best way to accelerate AI exploration is to look at AI pioneers. They’ve already done the homework – they have answered the
when it may not be needed.
questions “Is it real?”, “Can we win with it?”, and “Is it worth doing?”
• An understating of how AI is aligned with the
• Each AI use case is problem-specific and therefore its ownership needs to originate with the business and most probably a single
goals and strategic objectives of your
department. organization. Info-Tech Research Group | 55
• A cross-functional team will need to be assembled to evangelize AI and execute a pilot within your organization. • A draft of your organization’s AI mandate and
AI is making impact in many industries
This section provides an overview of the many uses of AI that are seen around the world today across many industries. As you read through the use
cases, think about the types of innovative projects that your organization has undergone recently and what challenges are still outstanding. Can the use
cases in this section serve as a starting point to answer your questions and to help achieve your organization’s strategic objectives?

Many industries are testing AI for potential uses. What is a use case?
Here we have included use cases for the following five
A targeted application of digital technologies to a specific
industries:
• Government • Education business challenge with a measurable outcome.
• Healthcare & Life • Manufacturing We recognize that use cases can be described at different
Sciences • Oil & Gas levels of granularity. The use cases mentioned in this
blueprint correspond to descriptions of specific business
challenges which practitioners in the industries and
Other industries include: sectors we studied acknowledged as meaningful and
• Banking & Financial • Retail impactful. There are many more use cases than what is
Services listed here. When making the selection, we strived to
• Gaming & Hospitality
combine depth (i.e. sufficient level of detail) with breadth
• Insurance • Media & Entertainment (i.e. variety of applications, processes, challenges) and
• Agriculture • Transportation proven business value. Our objective with this selection
• Construction • Telecommunication is to get your creative juices flowing.

Info-Tech Insight
Be sure to flip through the use cases from other industries. You may get some additional ideas and useful tools. For example, Dr. Atul Gawande
transformed intensive care by borrowing a take-off list from the airline industry and applying it to medicine. (Source: The New Yorker, 2007)
Info-Tech Research Group | 56
AI-Augmented
Government

Info-Tech Research Group | 57

AI use cases in Government
In Government, AI is being implemented and/or piloted to help achieve the following two broad objectives:
1) Increase automation and internal efficiencies – obtain insights to drive more informed decisions.
2) Improve citizen engagement, experience, and services.
The two objectives are obviously interlinked – automating repetitive, routine tasks helps to reduce backlog, free up government workers’ time to do more
complex things, serve citizens’ needs more quickly, and deliver more with the same resources. Mining data from complaints, suggestions, and other sources
helps to uncover ideas and insights for new services and policy changes. Below are some established and emerging use cases for AI and machine learning in
municipal, state/provincial, and national/federal government. The list is not exhaustive. Where information is available, we’ve listed observed and/or
anticipated benefits from implementing AI.
Use Cases Details/Examples Benefits
Chatbots/Virtual digital • Contact center/311-service chatbots help the public and businesses • The City of Los Angeles CHIP chatbot (created by two developers in just
assistants pay parking fines, understand tax rules, apply for a business three days) reduced email traffic by 50% by answering questions thereby
license, obtain permits, report potholes or other lapses in city helping to save internal resources.4
service, etc. • Easier, 24/7 access to information for citizens.
• Answer immigration-related questions and direct to the right
resources.1
• Internal bots help new hires with information about forms,
discussions, and the ever-present government jargon.2

Bots to automate applications DoNotPay bot that was initially designed to provide legal assistance • Simplified application process for vulnerable residents who may not be in
for government services to appeal parking tickets now helps newly homeless to file for the state of mind to fill out applications.
government housing and maximize their chances to get placed, for • Similar applications when created by governments can streamline data
free.3 collection and processing, eliminating manual work, improving internal
efficiencies, and delivering services to citizens faster.

Sources: 1U.S. Citizenship and Immigration Services n.d; 2Medium, “4 Government Agencies
Using Bots,” 2017; 3The Washington Post, 2016; 4cFocus, 2017 Info-Tech Research Group | 58
AI use cases in Government (cont’d)
Use Cases Details/Examples Benefits
AI for public health • AI symptom triage bot answers NHS 111 inquiries from more than a million • Provides access to care allowing people to avoid unnecessary
Londoners.1 doctor’s appointments; frees physicians on call to tend to
• App using ML and NLP that mines Twitter for food poisoning-related tweets residents in greatest need.
and generates a list of highest-priority locations for inspections. • More citations for health violations than the current random
• Access to care ‒ create a better understanding into where larger populations of approach, improving efficacy of inspections and creating fewer
elderly people and children are and appropriately distribute to health care food poisoning incidents and hospitalizations.5
facilities.
New policy development By correlating data between public housing and mental health, a local • Better services for citizens, such as for those with dementia in
government is looking to gain insights to update current policies and create new need of housing after treatment.
ones.
Public safety • Predictive policing; crime detection in real time; robocops. 2 • A 22% drop in residential burglaries after adopting a system
• Pre-trial release and parole decisions. that highlights possible crime hotspots where the police should
consider patrolling more heavily.6
Disease surveillance • Real-time surveillance that can expedite detection of outbreaks.  •  GPHIN detected Canada’s 2004 SARS outbreak two months
• Contact tracing. ahead of the WHO.7
• Public Health Agency of Canada’s Global Public Health Intelligence
Network (GPHIN), a digital, automated surveillance tool that collects
information from news feed aggregators and reports to national and
international health agencies.3
Expedited processing of patent Patent classification, image search, machine translation, determining • Increased data quality, processing speed, etc.
applications patentability, etc.4
Smart cities Using ML with satellite data to understand changing land use to enforce zoning Improve urban-planning governance; identify inconsistencies and
governance. potential violations.8

Sources: 1Digital Health, 2017; 2Mashable, 2017; 3Government of Canada, n.d.; 4WIPO, n.d.; 5National Science Foundation, 2016; 6The News Tribune, 2015; 7Global Health Now, 2018; 8World Info-Tech Research Group | 59
Bank Blogs, 2018.
AI in Healthcare & Life Sciences
"AI and Healthcare:
Cure-All, Poison 
Pill, or Simply
Smarter Medicine?"
– Healthcare Innovation, 2019

Info-Tech Research Group | 60

 AI use cases in Healthcare & Life Sciences
Use cases in Healthcare & Life Sciences fall into three categories:
1) Using AI and ML to advance medical knowledge by analyzing relationships between prevention or treatment techniques and patient outcomes.
Ranging from detection, diagnosis, and treatment protocol development to drug development, personalized medicine, and patient monitoring and care, the
objective is to create better-targeted medicines and treatment protocols that will result in optimal patient outcomes.
2) Digital consultation/virtual nurses – These patient-facing apps consider a person’s medical history and common medical knowledge to recommend an
action based on the symptoms reported (e.g., whether a doctor visit is needed). 1
3)  AI solutions for healthcare providers such as hospitals to improve internal operational efficiencies, decrease costs, optimize resource utilization,
improve patient satisfaction, and satisfy internal staffing and workforce needs.
Below are some established and emerging use cases for AI and ML in Healthcare & Life Sciences. The list is not exhaustive. Pfizer, for example, has over
100 projects across the company (in pilots or in production) that employ some sort of AI technologies (Davenport, 2019). We have listed observed and/or
anticipated benefits from implementing these use cases where available.
Use Cases Details/Examples Benefits
Identification and E.g., in immuno-oncology, an emerging approach to cancer Potentially shrinking development cycle (currently takes up to 12 years)
development of new treatment that uses the body’s immune system to help fight cancer and making it cheaper.3
drugs/drug classes or to develop drugs to contain disease outbreaks.

Telehealth/ Monitoring a patient’s symptoms via wearable devices and Empatica’s Embrace algorithm detects 93% of seizures.4
telemedicine alerting him/her and/or caregivers – e.g., about possible epilepsy
seizures.
Wearable robots From assisted limbs to exosuits and exoskeletons, 2 AI and robotics Wearable robots help with rehabilitation and movement support,
help paralytic individuals move. improve patients’ quality of life, and transform physiotherapy and
mobility industries.
Sources: 1Intertesting Engineer, 2019; 2PwC, “What Doctor?” 2017; 3Atomwise, 2015; 4Empatica Support, 2019
Info-Tech Research Group | 61
AI use cases in Healthcare & Life Sciences (cont’d)
Use Cases Details/Examples Benefits
Diagnostics • Interpreting imaging data (e.g., X-Rays, CT scans, images in • AI finds 95% of melanomas compared to 86.6% for
radiology, dermatology, and cardiology), thereby freeing up dermatologists.4
specialists’ time to focus on more complex cases. • First FDA-approved AI-powered medical device to detect certain
• AI analyzes millions of EMRs to identify clinical markers for diabetes-related eye problems.5
a rare form of heart disease. This data is then used to educate • Fast 3D-scan analysis can help surgeons determine early cancer
physicians on what to look for. surgery success.6
• AI-based blood test developed to identify cancer at the early • Early rare heart disease markers identify patients at risk.7
stages. • Face-scanning app spots signs of rare genetic disorders that aren’t
apparent to doctors.8
• Early cancer identification is expected to aid in better treatments. 9

Clinical decision-support systems • Identifying at-risk patients and medical emergencies such as An algorithm detects cardiac arrests with a 93% success rate
patients likely to suffer a cardiac arrest. compared to 73% for human 911 dispatchers.10
• Predicting whether a chronic wound will heal.1

AI in surgery • In orthopedic surgery, AI-assisted robotics analyze data from • AI-assisted robotic technique resulted in a five-fold reduction in
pre-op medical records to physically guide the surgeon’s surgical complications compared to when surgeons operated
instrument in real-time during a procedure. alone.11
• Cardiac surgeons use 3D-printed digital twins of patients’ • US $40 billion potential benefits by 2026 from robot-assisted
hearts to run simulations.2 surgical procedures like cardiothoracic, gastrointestinal, and
neurosurgery.12
Virtual nurses Avatars that interact with patients, ask questions about their HBR estimates that AI-powered nurse assistants could save $20
health, assess symptoms, and direct them to the most effective billion annually by saving 20% of the time nurses spend on patient
care setting. maintenance tasks.13
VR for autism VR helps people on the autism spectrum to sharpen social skills. 3 Reduced anxiety and fear, better social skills, improved integration
into society and employability.

Sources: 1Swift Medical, n.d.; 2Daily Mail, 2018; 3The Dallas Morning News, 2018; 4The Guardian, 2018; 5FDA News, 2018; 6Engadget, 2018; 7Biopharma Dive, 2018; Info-Tech Research Group | 62
8
Nature, 2019; 9Verdict Medical Devices, 2018; 10Bloomberg, 2018; 11Harvard Business Review, 2018; 12Accenture, 2017; 13Harvard Business Review, 2018
AI use cases in Healthcare & Life Sciences (cont’d)
Use Cases Details/Examples Benefits

Medical records and data • Robots collect, store, re-format, and trace data. • Faster, more consistent, and comprehensive patient data.
management • AI converts prescriptions into EHR entries. • Elimination of manual error-prone work.
• AI decodes medical records.1

Back-office administrative Activities that have nothing to do with patient care AI-based technologies can improve administrative workflows
activities consume over 51% of a nurse’s workload and nearly 16% and eliminate time-consuming non-patient-care activities such
of physician activities. as writing chart notes, filling prescriptions, and ordering tests,
saving the industry $18 billion annually.2

Medical error and fraud AI sifts through data to highlight mistakes in treatments More efficient processes and reduction of unnecessary patient
detection and workflow inefficiencies. hospitalizations.3

Carebots Robots designed to assist, monitor, and offer Deployed in tandem with human care workers, carebots can
companionship to the sick, the elderly, the dying. improve patient end-of-life care and prevent caretaker burnout.4

Sources: 1GeekWire, 2018; 2Harvard Business Review, 2018; 3Novatio, n.d.; 4NUVO, 2019

Info-Tech Research Group | 63

Case Study INDUSTRY SOURCE

Health Care University Hospitals, 2018

University Hospitals uses AI and ML to
improve its emergency department

Challenge Solution Results

University Hospitals Cleveland (UH) faces the UH turned to Hospital IQ’s Census Solution ED boarding hours have declined by 10%
same challenge that every major hospital to proactively manage capacity, staff, and and the hospital has seen a 50% reduction in
confronts regarding how to deliver increasingly flow in the ED and inpatient areas. the number of patients who leave the
complex, high-quality healthcare to a diverse hospital without being seen.
By applying AI, ML, and external data (i.e.,
population efficiently and economically. In 2017,
weather forecasts) to the hospital’s own data UH also predicts in advance patients ready for
UH embarked on a value improvement program
(including EMR data and hospital policies), discharge and identifies roadblocks, reducing
aiming to improve quality while saving $400
the solution helped UH make two-day census the average LOS by 15%. UH is able to better
million over a five-year period.
forecasts that managers used to determine manage staff, reducing overtime and cutting
In emergency department (ED) and inpatient whether to open or close in-patient beds and, overall labor costs.
units, leaders found anticipating demand difficult, when necessary, divert low-acuity patients to
The hospital has also increased staff
and consequently units were often over-staffed other hospitals in the system to handle
satisfaction and improved patient safety by
when demand was low and under-staffed when predicted patient volume.
closing specific units on weekends and
demand was high. Hospital leaders were uncertain increasing the number of rooms that can be
about how to reallocate resources based on sterilized.
capacity needs. Info-Tech Research Group | 64
AI in Education

"Intelligent machines will replace teachers within

ten years."
Info-Tech Research Group | 65
 (Source: Independent, 2017) 
AI use cases in Education
Use Cases Details/Examples Benefits
Intelligent instruction • Capabilities range from textbook summarization, to • Better, more efficient learning experience.9
design/personalized tutoring, custom content creation and textbooks design,1 to • Identification of students who struggle so that they can
adaptive learning personalized guidance, assessment, and testing.2 get help faster.
• Based on social media, grades, and marks determining
students who struggle to help them.

Feedback to tutors on lesson Real-time feedback to get a tutor back on the topic or to try a More engaged and effective instruction.
plans different approach if a student is disengaged.3
AI for universal access Apps that help students with visual or hearing impairments • More inclusive access to education.
by narrating the world around them,4 translating presentations • More diverse classrooms.
in real-time, etc.5
AI scoring and grading AI helps teachers to grade essays, recommending • Teachers can spend more time with students.
homework improvements (e.g., writing style, structure, and theme).6 • AI helps teachers avoid inconsistencies caused by
human errors such as lapses in attention or unconscious
bias.10

AR and VR to improve learning • At The University of Arizona, students are being taken to a • Immersive experience leading to better learning.
virtual Harlem as well as studying phobias using virtual • More accessible education.
spiders.7
• VR may help learners with autism.8

Sources: 1CTI, n.d.; 2Derandomized, 2019; 3Medium, “The Role of AI in Education,” 2018; 4Microsoft, n.d.; 5Microsoft, 2018; 6South China Morning Post, 2018; 7EdSurge, 2018;
8
The Dallas Morning News, 2018; 9Carnegie Learning, n.d.; 10South China Morning Post, 2018.

Info-Tech Research Group | 66

AI use cases in Education (cont’d)
Use Cases Details/Examples Benefits

Performance-based career/ AI systems help students to select majors based on where • Majors and future careers closer aligned with
college major advice they succeed and struggle.1 students’ interests and skills.
• Fewer drop-outs.

Robo-tutors/tutor-bots More than a glorified FAQ, robo-tutors may learn to ask • Personalized curriculum; 24/7 access.
questions and engage students in a dialogue to improve • Cheaper education for students.
learning and offer a more meaningful interaction.2 • Cost savings for the educational institutions, life-
long learning, customer loyalty, and revenues.

Sources: 1TeachThought, 2018; 2The Economist, 2018

Info-Tech Insight
One of the objectives of educational institutions is to prepare students for future work (which may or may not be coding). So
educators need to actively track and anticipate labor market trends as they pertain to AI and skills that are/would be in demand.
AI is changing the nature of jobs. Like many technologies before it, AI will replace some occupations and it will also create new
job categories.

Info-Tech Research Group | 67

AI in Manufacturing and Construction

Info-Tech Research Group | 68

  
AI and digital twin are made AI and ML are being used in
60 percent of IoT activities.
for each other ‒ McKinsey & Company, 2019

Digital twins are virtual replicas of physical objects, devices, people, places,
processes, and systems. In Manufacturing, almost every product and
manufacturing process can have a complete digital replica of itself thanks to
IoT,* streaming data, and cheap cloud storage.

All this data has allowed for complex simulations of, for example, how a
piece of equipment will perform over time to predict future failures before
they happen, reducing costly maintenance and extending equipment lifetime.
In addition to predictive maintenance, digital twin and AI technologies
have enabled organizations to design and digitally test complex equipment
such as aircraft engines, trains, offshore oil platforms, and wind turbines
before physically manufacturing them. This helps to improve product and
process quality, manufacturing efficiency, and costs.

*For more insights and

Understand and Apply Internet-of-
use cases on IoT beyond
Things Use Cases to Drive
manufacturing, see Info- Organizational Success
Tech’s blueprint

Info-Tech Research Group | 69

LINAMAR is piloting the Factory of
the Future
INDUSTRY SOURCE
Manufacturing Creative Destruction Lab, 2018

Case Study
LINAMAR manufactures and supplies automotive and industrial
markets with factories across North America, Europe, and Asia. It has
59 manufacturing plants, six R&D centers, and 21 sales offices in 17
countries. The company is piloting a “lights-out” manufacturing
facility where it can fully automatically run equipment using
autonomous lift-trucks, tools changers, and robotics to run these
machines.

"This is in a pilot stage where it’s been

encountering all kinds of challenges but
also all kinds of interesting opportunities
and learning a lot in the process as
well."

‒ Linda Hasenfratz, CEO

(Source: Creative Destruction Lab, 2018)
Info-Tech Research Group | 70
 LINAMAR uses machine learning to
innovate
INDUSTRY SOURCE
Manufacturing Creative Destruction Lab, 2018

Case Study
LINAMAR is exploring ML and AI for:
•Product innovation
Product Development
•Process innovation MaterialDigital
Tangential
Process
Product
Manufacturing Through
Development
Innovation
AI/ML
Development
Development
Partnerships
– Lighter,
WithStronger
Startup
•Material development
•Tangential innovation (i.e., net-new products)
•Manufacturing partnerships with tech start-ups
Digitization With Machine Learning
•Digitization through AI/ML 1,124 LMMS
2,189 Connected
60 Plants Data Collection
Machines
Connections

"We have always believed that our

competitiveness as an organization absolutely 1,164 RFID 1,788 Traceability 823 Vision
Stations Read Stations Systems
hinges on our ability to be innovative and to,
obviously, be efficient in how we are running our
operations. And we see machine learning as a key
851 Traceability
element in achieving both of those different 2,590 Robots 2 AGVs
Marking Stations
areas."
Info-Tech Research Group | 71
‒ Linda Hasenfratz, CEO
 Integrate Artificial Intelligence Into Your
For more examples of AI use Manufacturing Operations
cases in Manufacturing, see Use this blueprint to learn how to easily identify the AI initiatives
that your organization is best positioned to implement and to
Info-Tech’s blueprint: learn how to build the business case for AI in your manufacturing
organization.

Info-Tech Research Group | 72

AI in Oil & Gas
Info-Tech Research Group | 73
AI in the Oil & Gas value chain

Exhibit: The First – SPE Norway Magazine, 2017. Used with permission.

Info-Tech Research Group | 74

How Oil & Gas companies are prioritizing their AI investments

Chemical and petroleum companies prioritize cognitive computing

across the value chain

Upstream exploring, appraising, drilling, and production optimization (such as field economic value, geo-political analysis) 15%

Regulatory compliance 26%

Asset management (such as plant maintenance, management, fleet/vessel/retail tanker management) 34%

Environment, health and safety 38%

Bulk distribution and retail 43%

Downstream production planning, trading, and risk management

Source: IBM, 2017 48%

Info-Tech Research Group | 75

Additional AI use cases in Oil & Gas
Use Cases Details/Examples Benefits
Processing plants that run on Sinopec, in partnership with Huawei, is developing ten By keeping oil in reserves as opposed to pumping it up to the
AI processing plants run almost entirely by Al. surface, Sinopec aims to reduce operation and production costs by
about 20%.1
AI robots for ocean ExxonMobil teamed up with MIT to design AI robots for ocean This can contribute to protecting the ecosystem and serve as an
exploration exploration that will be able to search the ocean to detect oil indicator for substantial energy resources.2
seeps.
AI mapping potential for oil Total has been undertaking hydrocarbon exploration, which is By using AI, Total is looking to find more innovative and
and gas deposits the ability to map and identify oil and natural gas deposits environmentally friendly methods of achieving this.3
beneath the earth’s surface.
AI to optimize shale extraction Shale wells decline quickly compared to conventional wells. By By shifting the decline curve for shale to more closely match that
using AI, engineers can better understand and model out of conventional wells (basically, lengthen the curve, and increase
permeability and porosity of the reserves and use that insight to the volume of output), petroleum engineers will add 5% to global
frack these wells differently. reserves.4
Seismic imaging Can be used to more easily identify deposits. Better and faster interpretation of seismic data to identify
hydrocarbon deposits.5
Asset optimization from wells Multiple use cases: production allocation and planning, gas lift Predict an asset’s future performance, recommend appropriate,
to processing facilities optimization, gathering network timely actions to improve production uptime and to optimize
(asset optimization and process optimization). operating condition.6
Automated royalty AI engine reads the complex JV contracts, pulls out the terms Processing time reduced from 800 seconds per royalty contract to
calculations using AI, RPA, that define the royalty calculation, feeds the results into an RPA three seconds, fault free.7 Reduced AP costs, elimination of
and blockchain tool that then converts the terms into code. That code becomes a disputes between royalty partners, and reduced stress and risk. 8
smart contract on blockchain.

Sources: 1WildFire, 2018; 2Emerj, 2019; 3Emerj, 2019; 4EnergyNow Media, 2018; 5The First SPE Norway Magazine, 2017; 6World Oil, 2018; 7EnergyNow, 2018; 8GuildOne, 2018 Info-Tech Research Group | 76
Step 2.2: Give Your AI a Purpose

PHASE 2

2.1 2.2
Understand How Your Give Your AI a Purpose
Peers Are Using AI

This step will walk you through the following activities: This step involves the following participants:
• Now that you understand what AI is, what technologies are • CIO, CTO
available for use, and how your peers are deploying AI, the next • IT executives and managers looking to learn about AI
step will help you determine the top three business goals to get • Analytics practitioners
value from your AI. • Project/program manager(s)

Outcomes of this step

• A list of business problems or opportunities that AI could potentially help solve within our organization.
• An understanding of when AI should be used, and when it may not be needed.
• An understating of how AI is aligned with the goals and strategic objectives of your organization.
• Drafting an AI mandate and charter, and identifying the team involved in this initiative.

Info-Tech Research Group | 77

Give your AI a purpose by creating clear objectives
Purpose. It is something that we are all trying to find. For a new AI, it is critical to give it a reason to
exist, and an end goal that can be achieved to showcase success.
This phase is focused on the these first two
steps. Your organization will not find value
from its AI initiatives until you narrow
down your goals to specific objectives with
which to prove value and showcase results.
It is imperative to define these objectives Design AI
before starting on the design phase because
the reality of the field of AAI is that it is Define
narrow and also rapidly changing and Objectives
evolving.
Ask
questions
Unless you put a defined scope around the
initiative at the outset, your AI initiative
may not succeed, or you will never be
“done” as you continue to chase new
promises.
Info-Tech Insight
The key question to ask is not “What can AI technologies do?” but rather “What can they
do for us?” Put differently: “How much would we benefit from AI if we were to invest in
it?”
Info-Tech Research Group | 78
Value-generating AI requires deep understanding of the business and the ability to critically
ask questions
Give your AI a purpose.
Although complex, the algorithms behind AI technology are not the most difficult part of data science and machine learning. The most difficult part of
leveraging AI technology is finding the right question to ask – and the right business problems to solve. (Not all problems can or should be solved
with AI.)
Most organizations start with AI by running pilot projects which look to increase internal efficiencies,
optimize processes, and decrease costs. This approach is familiar from deploying analytics. Starting small
with internal cost containment focus is smart. It allows one to learn the new technologies; better understand Info-Tech Insight
the dependencies, opportunities, and fit with your organization; minimize risk; and deliver value quickly 1. Learn from AI pioneers:
(and on a small budget).
Start small with AI, but keep
However, as organizations become more experienced with AI, their focus shifts on the potential for in mind that the ultimate
growth – using AI to expand the business by creating new products or services, increasing productivity, or objective and value comes
winning more market share (MIT Sloan Management Review, “Five Management Strategies,” 2017). from growing the business,
Ultimately, you will need an AI strategy closely aligned to the strategic goals and drivers of your not cutting costs.
organization (MIT Sloan Management Review, “Artificial Intelligence in Business Gets Real,” 2018).
2. While working on tactical
Drivers to Establish Competitive Advantage pilot projects, start thinking
about your AI strategy to
Customer Intimacy/ Product Innovation/ Operational
Risk & Compliance
catch up to the early adopters
Service Excellence Service Excellence and get ahead of the
competition.

"Stay true to who you are as a business and then innovate on top of that
with new technologies and new things. Speak to the reality of what is ‒ Justin Reilly, ex-Head of
CX Innovation, Verizon
your mission and what do human beings need from you to be happy with (Source: AIIA Network,
Info-Tech Research Group | 79

you as a brand." 2019)

What is your AI strategy? What are you looking to achieve and why?
AI technologies can support many different business objectives, and so it is important to decide which ones to focus on. This decision should be
driven by your organization’s business strategy; it may evolve over time as the organization becomes more familiar with the technologies and
starts realizing their value. Below are a few examples from AI pioneers. Source: The AI Advantage, 2019

Verizon initially focused on externally- GE has both internal and external Nasdaq’s key focus for AI has
oriented application of AI that could lead focus. Internally AI is used, among been on identifying potentially
to market growth: other things, to unify and integrate fraudulent or illegal financial
• Recommendations for new products key data elements, such as suppliers traders and trades. It applies these
and services. and customers, across its many capabilities internally and also
• Intelligent agents. databases and systems. (It had, for offers them to other stock
• Customer analytics to identify revenue example, over 50 different version of exchanges and to buyers and
growth opportunities, etc. CRM.) sellers of financial instruments.
Today, it’s also using AI to improve Externally it uses digital twin and AI This is an example of using AI to
network performance (speed and quality of for predictive asset management of advance external regulatory
its service) to ensure that customers its industrial devices. This is a compliance and risk management
receive what they are paying for (so, product-oriented approach, as this objectives. (As well as an
internal focus). makes GE’s products and services example of product/service
more competitive. innovation.) The AI Advantage, 2019
Forbes, 2018 The AI Advantage, 2019

"As with most strategies, the goal of a cognitive strategy is to surface, answer, and achieve consensus on key questions as an
organization." Info-Tech Research Group | 80
– Davenport, The AI Advantage, 2019.
Successful execution of an AI strategy requires careful
change management and putting humans first

People, not technology, are key to any technology-driven

transformation.
Your vision for AI will need to be supported by other components for you to
successfully shepherd the organization through its AI-enabled transformation. (Some
would say “imposed,” and we think it’s both.) Transformation is change, and change
is:
CHANGE = VISION + SKILLS +
INCENTIVES + RESOURCES + ACTION
PLAN
Unless you have all five components defined and in place, you will encounter confusion, anxiety,
resistance, frustration, and false starts.
Minimize these issues by doing the following:
• Put in place a plan to close gaps in skills.
• Define “what’s in it for me” for all levels of the organization.
• Ensure sufficient resources are available to achieve your vision (that means having a budget).
• Create a roadmap and an action plan (short- and long-term).
• Have a communication and engagement plan and communicate clearly and frequently.
Info-Tech Research Group | 81
To ask the right questions of your AI, you need business
understanding and context
The first step in leveraging AI technology is to determine a set of projects in which AI can potentially be
used.
Info-Tech Insight
Business pressing issues Define Prioritize Can AI solve Create AI An AI use case is problem
questions questions this problem?  objectives specific and therefore its
Business goals
ownership needs to originate
Find enthusiastic, respected champions in the business who have pressing initiatives from the business and most
that address one of the business goals below and partner with them to determine how probably a single department.
AI will bring value to their goals.

Business Goals for AI

Automation Integration Scale Insight/ Impact/

The ability to be lean The ability of AI The ability for the AI Intelligence Engagement
through the replacement of technology to integrate product and/or service to The ability for the AI The ability of the AI
aspects of human-based with the technology of quickly increase capacity. technology to leverage technology to deliver a
organizational processes customers, partners, and organizational outputs to positive social impact on
and activities with AI suppliers. make processes smarter user and customer
technology. and more efficient. experience.

Info-Tech Research Group | 82

AI isn’t just about data and algorithms. Assemble a multi-disciplinary, business-
led team to explore the possibilities
2.1 Duration: 1-2 hours

Instructions The Project Lead has:

At this stage, start thinking about the highest-value application of AI technology to your • Vocal interest in innovation
organization. AI use is department- and problem-specific, and it should be treated as such. Identify • Analytics experience
which department is a good candidate for your first AI pilot and a business person to sponsor this • Strong communication skills
project. This person should be open to change and experimentation and not afraid of failure. • Extensive experience translating between
(Your first AI project may not be a complete success.) business and IT
You will also need an executive sponsor (CXO level) to set a compelling strategic vision • Strong relationships with the lines of
for AI and champion the initiative, and a Project Lead to lead it. The project lead will: business
1. Work with the Executive Sponsor to define and document the AI mandate. • Program and project management
experience
2. Assemble and chair an AI working group.
Team members to consider:
3. Lead the effort to identify opportunities to leverage AI in your organization, assess and • Data owner for key data domain – aka
prioritize ideas, and facilitate prototyping. your business sponsor
• Data steward(s)
Info-Tech Insight
• Chief architect/Enterprise architect
The data architecture function is innovative by design, tasked with understanding the current and future
• Director/VP of applications
state of the organization. If you have a strong staff member in this practice, they may be a good candidate • Data scientist/Data analyst
to lead AI. They also typically speak both the language of technology and the language of business. • BI/Analytics developer
• DWH architect
• Data integration lead
Depending on the resources and skillsets available, building your AI team may involve external hiring. • Senior BA
Consider also leveraging vendors and third parties. For more details, see Info-Tech’s Industry note • External partners Info-Tech Research Group | 83
No Data Scientist? – No Problem; You Can Still Get Your Feet Wet With AI!
 will
Looking ahead: AI-related skills you may need ASAP
Data Scientist: Turns raw data into insights and uses statistics, machine Data/Business Analyst: Captures requirements; analyzes data to tell a story
learning, and analytical approaches to solve business problems; has and to produce actionable insights; creates and disseminates charts, models,
strong programming skills, ability to create new algorithms, knows how visualizations; presents analyses to stakeholders who make business
to handle big data, and has expertise in one or more business domains. decisions.
Applied Machine Learning Engineer: Applies existing AI algorithms Data Engineers/DataOps team: Builds production pipelines that supply
on your organization’s data; iterates to achieve the desired results, data scientists and business users with data (Eckerson Group, “DataOps,”
although this category may disappear quickly, given the trend of 2018).
automating training and pending further developments in transfer AI/Data Science Manager: “A force-multiplier, ensuring that your data
learning (Hackernoon, 2018). See, for example, Google’s AutoML. science team isn’t off in the weeds instead of adding value to your business”
(Hackernoon, 2018).

Info-Tech Insight

Most likely you already have

internal talent you can nurture into
some of the above roles. See
No Data Scientist? – No Problem;
You Can Still Get Your Feet Wet Wi
th AI!
on how to get started.
You will also need to engage your
DevOps team and other IT
resources to be able to run AI
reliably at scale and within SLAs
Info-Tech Research Group | 84
and to deliver the last mile.
Exhibit: SAS Blog, 2013. Used with permission.
Looking ahead: AI-related skills you may need
long-term
• Machine relations manager (like HR Managers): Oversees AI systems, conducts performance
reviews, promotes systems performing well, replicates and deploys to other parts of the
organization, demotes/decommissions underperforming AI systems.
• AI trainer: Trains AI to capture empathy, personality, worldview, and localization differences.
• Interaction modeler.
• Data hygienist: Removes biases.
• Algorithm forensic analyst: Helps explain the inner workings of algorithms – e.g. why credit
was denied.
• Transparency analyst: Classifies and maintains information about AI system’s accessibility OR
(because many AI algorithms are a black box).
• Explainability strategist: Decides which AI technology may be deployed for specific
applications.
• Context designer: When developing a new system, assesses business environment, processes,
tasks, users, cultural issues, etc. to ensure AI-powered systems are designed right. 
• AI safety engineer: Anticipates unintended consequences and addresses any harmful
occurrences. 
• Ethics compliance manager. 
• Automation ethicist: Evaluates non-economic impact of AI systems. Source: Human + Machine, 2018
Info-Tech Research Group | 85
BAs and PMs are your key resources and secret weapon for starting
a successful AI movement in your organization

Most urgently, get your people skilled up, in particular Business Analysts (BAs) and Project Managers (PMs)/Product
Owners (POs). 
They will be your key resources and secret weapon for a successful AI initiative. Watch the three-minute video by Derek Sivers “
"The firstis.follower is what transforms
First Follower: Leadership Lessons from a Dancing Guy" on how important the first “follower” a lone nut into a
leader.
           New follows emulate the followers, not the
leader."  
‒ Derek Sivers

Info-Tech Insight
Your Business Analysts and Project Managers/Product Owners can and
should be your first followers to help you start the AI movement in your
organization.

For more information on who Product Owners are and what they do, see
Info-Tech’s Product Owner job description.

Image: Derek Sivers, sivers.org, used with permission.

Info-Tech Research Group | 86
Create the mandate of your AI working group and charter
2.2 Duration: 1-2 hours

The mandate is intended as a quick document to guide the team to the problems and opportunities that are Participants
most critical to the business and are solvable with AI. • AI Business Sponsor
The charter will help to keep the team on track. This should be a light-weight document explaining the • AI project lead
objectives of the initiative and the rules of engagement. It should contain the following: • Data steward(s)
• Purpose (mandate) of the initiative • Chief architect/Enterprise architect
• Who will be involved; their roles and responsibilities Give your AI working group a • Director/VP of applications
• Activities to be performed memorable name. People are more • Data scientist/Data analyst
• Meeting cadence likely to engage and stay engaged if • BI/Analytics developer
• Deliverables they are working on an engaging
• DWH architect
project. (Yes, that was intentional.) Be
• How you will measure the group’s (and organization’s) success creative! • Data integration lead
• Senior business analyst
Instructions
1. Draft a mandate on a whiteboard or a flip chart as one large group or by splitting the team into smaller groups Materials
to encourage friendly competition.
• Areas of focus identified in
High-level mandate Areas of focus Exercise 2.2
This should describe what you want to achieve with These are the activities, processes, and services • Whiteboard/flip chart
AI, linking to your organization’s strategic that present the greatest opportunity for AI. • Markers
objectives.
2. Discuss the above and the rules of engagement. Finalize. Having achieved consensus, document in the group
charter.
Info-Tech Research Group | 87
3. Review the charter with the Executive Sponsor and obtain approval.
Brainstorm opportunities for achieving your specific business goals with AI

2.3 Duration: 2-3 hours

Instructions
With your AI team, hold a brain-storming session where you ask the following questions:
INPUT
1. What are your most pressing
business problems and • Your organization’s
"AI shines in problems where the goals  strategic plan
opportunities?
are understood, but the means aren’t." • AI use cases in Step 1
2. What are your most important – Medium, “Asking the Right Questions About AI,” 2017 of this blueprint
business objectives?
3. How do they correlate with OUTPUT
known AI use cases and Increasing
success stories (in Step 1)? Goals Environment • Opportunities to use AI
difficulty of
4. For new ideas not covered in applying AI, Clear Predictable
earlier use cases: rate these as unknown
Clear  Unpredictable
Materials
value and fit
“AI-able” or not using the • Whiteboard, markers
Indirect Predictable Decreasing
framework to the right. (More novelty of AI, for answers
details and examples are on the Indirect Unpredictable established
next two slides.) value, proven Participants
Undefined Any
5. Document your ideas. technologies
• AI project team
Focus here

To ensure success in your new AI initiative, the project must show value to the business while also addressing these criteria:
1. The problems that AI will solve require prediction of outcomes, not understanding what causes something else (causal
inference).
2. The problem needs to be well-understood and self-contained.
Info-Tech Research Group | 88
Where AI can help and where it can’t (yet)
AI is about making predictions about possible outcomes.
Let’s re-cap: AI is, in essence, a prediction machine. When applied to data, an ML algorithm extracts a number of features and
creates a model, i.e. a set of possible outcomes and their probabilities. When this model is next applied on new data previously
unseen by the algorithm, it predicts the likelihood of these outcomes for the data points in this new data set.

Historical
1 Algorithm Model
data

2 New data Model Predictions

So, which problems are AI-able?

Goals Environment Examples of Activities Suitable for AI?
A camera looking at the assembly line expecting a Yes, these tasks have been automated for several years, with good
Clear Predictable success.
car to show up.
Clear  Unpredictable Driving a car. Getting better at this (e.g. self-driving vehicles).
Learning to play a game (hard, because the Yes, suitable for AI and good progress has been made (e.g.
Indirect Predictable relationship between actions and consequences may AlphaGo).
be distant).
Indirect Unpredictable Planning your financial portfolio. No progress yet, but maybe over the next decade.
Not suitable for AI unless broken into sub-parts or redefined more
Undefined Any Writing a novel. narrowly (e.g. “write a novel which will sell well if marketed as
horror”).
Info-Tech Research Group | 89
Source: Medium, “Asking the Right Questions About AI,” 2017.
When blueskying ideas, focus on small projects that will deliver business outcomes

When thinking about how and where AI can help you, start with asking these questions:
• Is your task or process data-driven?
• Do you have suitable and sufficient data to support an AI-driven solution?
• Do you understand how AI works?
• What do you want AI to do? Is there a clear outcome or goal for the task or problem you Consult Info-Tech’s
are considering? Can you define what success (and failure) look like? – You can’t train a blueprint
system if you can’t tell it what to do. Kick-Start IT –Lead Bu
siness Innovation
• Is the data you’ll use to train your AI representative of the problem you want to solve? for methodology and
Is it relevant, so that AI stands a chance to learn from it? tools to help ideate,
• Is there a limited number of options your AI will need to consider? It should not be an define, refine, prioritize,
overwhelming range. (Because, what’s the point?) and prototype your
• Yet is the problem complex enough that you can’t just write down the rules to solve it? ideas.
(Because, otherwise, what do you need AI for?)
• Can you connect predictions to business actions and outcomes – and can they be
measured? – so that you can see what did and didn’t work?

"It might be tempting to start your AI transformation by trying to completely reinvent your business
process. However, this is the best way to fail. It is wiser to gain experience from smaller projects that
will create a solid foundation to move onto bigger, more complex human interactions and AI
challenges." 
Info-Tech Research Group | 90
– Alexandre Gonfalonieri, “What I Have Learned After Several AI Projects,” Medium, 2017
 This phase will walk you through the

Phase 3
following activities:

3.1 Develop Your AI PoC Project Ideas

• Contextualize AI peer use cases for
Select Your First AI Proof of Concept your organization.
• Create problem statements.
• Prioritize and sequence your AI ideas.
Get Started With Artificial Intelligence • Select your first AI PoC project.

3.2 Create an MVBC for Your First AI

PoC

• Create an MVBC for the selected AI

use case.

Phase 1 Phase 2 Phase 3 This phase involves the following

1.1 Understand AI Trends 2.1 Understand How Your 3.1 Develop Your AI PoC participants:
and Technologies Peers Are Using AI Project Ideas
• AI working group.
1.2 Understand 2.2: Give Your AI a Purpose 3.2 Create an MVBC for
Organizational Impact of AI Your First AI PoC Phase Outcomes
• Several AI use cases have been
identified as being of value to your
organization. Top idea has been
selected for a PoC implementation.
• All other ideas are documented in the
Idea Reservoir Tool and have an
Phase 3 Results & Insights: owner.
• Several AI use cases have been identified as being of value to your organization. Top idea has been selected for a PoC • A strong MVBC for your first AI
implementation. project has been created and
• All other ideas are documented in the Idea Reservoir Tool and have an owner. If no one champions an idea, it’s not worth submitted for approval by the
pursuing. Executive Sponsor and the
• A strong MVBC for your first AI project has been created and reviewed and approved by the Executive Sponsor and the governance bodies.
Info-Tech Research Group | 91
governance bodies.
Phase 3 outline

Call 1-888-670-8889 or email [email protected] for more information.

Complete these steps on your own, or call us to complete a guided implementation. A guided implementation is a series of
two to three advisory calls that help you execute each phase of a project. They are included in most advisory memberships.

Guided Implementation 3: Select Your First AI Proof of Concept

Proposed Time to Completion: 1-2 weeks

Step 3.1: Develop Your AI PoC Project Ideas Step 3.2: Create an MVBC for Your First AI PoC
Start with an analyst kick-off call: Review the results of your brainstorming and project selection
with the analyst:
• Review the methodology for contextualizing peer use cases
for your organization. • Discuss problem statement and expected value.
• Discuss pre-requisites and dependencies.
• Assess feasibility.

Then complete these activities… Then complete these activities…

• Conduct brainstorming with your AI team. • Create a minimum viable business case (MVBC) for the selected
• Discuss and prioritize projects. use case. Review with the Info-Tech analyst.
• Document ideas in the Idea Reservoir Tool. • Submit to your Executive Sponsor and relevant governance bodes
• Select your first AI PoC project. for review and approval.

With these tools & templates: With these tools & templates:
Problem Statement Template Minimum Viable Business Case Document
Idea Reservoir Tool Prototyping Workbook

Phase 3 Results and Insights:

• Several AI use cases have been identified as being of value to your organization. Top idea has been selected for a PoC implementation.
• All other ideas are documented in the Idea Reservoir Tool and have an owner. If no one champions an idea, it’s not worth pursuing.
• A strong MVBC for your first AI project has been created and reviewed and approved by the Executive Sponsor and the governance bodies.
Info-Tech Research Group | 92
Step 1: Develop Your AI PoC Project Ideas

PHASE 3

3.1 3.2
Develop Your AI PoC Create an MVBC for Your First
Project Ideas AI PoC

This step will walk you through the following activities: This step involves the following participants:
• Contextualize AI peer use cases for your organization. • AI working group.
• Create problem statements.
• Prioritize and sequence your AI ideas.
• Select your first AI PoC project.

Outcomes of this step

• Several AI use cases have been identified as being of value to your organization. Top idea has been selected for a PoC implementation.
• All other ideas are documented in the Idea Reservoir Tool and have an owner.

Info-Tech Research Group | 93

Contextualize selected peer use cases for your organization

3.1 Duration: 1-2 hours

Instructions
For each of the AI use cases in Phases 1 and 2 that resonated with the team, answer the following 12 questions to INPUT
further develop the opportunity and create a problem statement. Use the template on the following slide to record
• Use cases from the
your answers. The problem statement should be clear and concise and need-based and should not jump to
solutions. The problems it describes should be contained, fixable, aligned with the organization’s strategy, and previous sections and
solving them should deliver clear value. the team’s expertise

The first three questions are “OR” questions; you need to answer just one of them for each idea. The last three
OUTPUT
questions dive into technical details; they will help assess the idea’s feasibility but they are not necessary to create
an MVBC. • Problem statements
1. What top business opportunity do you wish to see realized with AI? (Hope) OR
2. What top risk will AI help you to alleviate? (Fear) OR Materials
3. What top pet peeve can you relieve with AI? (Frustration)
• Note-taking materials
4. What would be the challenges if performing this activity or process today?
• Whiteboard or flip chart
5. What would it look like if we could solve this perfectly? • Markers
6. What other business activities/processes would be impacted/improved if we solve this?
7. What compliance/regulatory/policy concerns do we need to consider in any solution?
8. What are the benefits of the new solution, and who will benefit from them?
Participants
9. What measures of success/change should we use to prove value of the effort (KPIs/ROI)? • Your AI team including
10. What are the steps for the process/activity? representatives from the
11. What are the applications/systems used at each step and from one step to the next? business units
12. What data elements are either created, used, or transformed at each step (high-level)?

Reframing the problem from the perspective of the end user is a very powerful technique. It will
prevent you from designing a solution that doesn’t address the real pain being experienced by
your target user.
Info-Tech Research Group | 94
Problem Statement Template: AI Idea One <Idea Name>
Problem Statement

What top business opportunity do you wish to see realized with AI? OR What top risk will AI help you to alleviate? OR What top pet
peeve can you relieve with AI?

Current Challenges What Does Amazing Look Like?

Related Processes/Impact Compliance Benefits and KPIs

Current Process & Systems Applications & Systems Data Elements Involved

Info-Tech Research Group | 95

Documenting a process (aka question number ten)

Don’t strive for perfection; a quick diagram can be very powerful.

When documenting a current or desired future process, don’t strive for perfection. A quick drawing on a whiteboard or a flip chart (or in
PowerPoint, Paint, or other e-tools) may be all that’s needed at this stage. The objective is to capture the process and reasons why it
should be replaced or optimized. Consider, for example, the drawings below from a member workshop.
The drawing on the left was all that was needed to convince the member executives of the value of the project. When they saw how the
data is currently being replicated to downstream systems and analytical applications rather than consulting the system of record
(enterprise data warehouse or alike), the project was deemed a “no brainer” and “a must.”
Remember: a picture is worth a thousand words.

Images source: Info-Tech workshop

Info-Tech Research Group | 96
Ensure every AI use case has an owner and a support group to prototype and test it

Every idea determined worthy of pursuing should have an owner. Optimally, owners are volunteers, and they
should self-select rather than being assigned. If no one is willing to own the idea, you should question its
value.

Use case owners should:

A
• Create and socialize a compelling business case D Define KPIs to measure success and impact.
for their idea. Present to executives.
Confirm the value/impact of the idea from prototype
• Participate and/or lead prototyping should this E testing. Ideally, some value should already be realized,
B
idea be selected for prototyping by the AI even during this stage.
working group.
F Help socialize any future iterations of the idea with
C
• Ensure all required resources are available (or users.
can become available) for prototyping and
testing.

Idea Support Group:

Have the idea owner identify two to three individuals with the
expertise and passion to participate in any future prototyping.
This group of individuals will make up the support group.

If an idea is not voluntarily picked by an owner, it isn’t worth pursuing further.

– Anonymous, former CIO of Royal Trust
(Source: Info-Tech interview)
Info-Tech Research Group | 97
Interlude: Are you ready for AI?
AI requires a certain level of maturity in data management, enterprise architecture, infrastructure, and
governance:
Do you have enough data to train your AI and how good is it? Do you have the appropriate infrastructure in place?
Data is what fuels AI. And you need a large quantity of data that is varied and is All that data needs to be easily accessible, often in real-time.
of a reasonable quality. • What does your big data architecture look like? Can it support your use cases and
• Do you have enough? (And is it digital?) beyond? Is it scalable? Flexible?
• What’s the quality of the data? Do you need to clean data, enrich it, etc. before • Which AI frameworks and tools will you be using? Will you build, buy, rent, or use
you feed it to AI? open source?
• If you don’t have the data you need, should you start collecting it? What • Where do they fit into your current enterprise architecture? (And do you have the
would be required to do that? How long will it take? Can you obtain it necessary skills?)
somewhere else? • Once you start embedding AI into business processes and applications to support real-
• How good is your master and reference data? time decisions, can you integrate with operational systems dynamically, in real-
time?
All of this takes time and effort, but good results cannot be achieved otherwise.
Where are you with governance?
For AI, you should at minimum consider governing the following:
Have you considered the “last mile”?
• Data
To deliver maximum value, AI-enabled insights need to be integrated with the • Models
business processes, systems, and tools that people use to make decisions and • Decisions
take actions every day. Can you answer the following questions when requested:
How successful have you been at embedding analytical content into intuitive • What data was used and where did it come from? (Lineage.)
workflows and business applications so that users don’t need to step out of • How was the decision made? How did the system arrive at a particular conclusion? (Most
whatever tool they are using to gain insights? They should be able to gain those
AI are black box.) Which model version was used? (Explainability, decision rationale,
and take actions in the same interface.
model history.)
It’s the same with AI. Can you deliver AI seamlessly to the point of decision • How is accountability divided between humans and machines? (Accountability and
making?
authority.)
Info-Tech Research Group | 98
To mature your data-management capabilities and prepare
for AI, use Info-Tech’s resources
Recommendations Click on the links below to access
Info-Tech’s resources.
1. Assess the maturity of your data management practices and understand
where the gaps are based on the AI use cases being considered.
• Are You Ready for AI?
2. Create a data strategy to ensure you have a steady supply of rich, • Leverage Big Data by Starting Small 
high-quality data to power near-term and future AI applications,
including what additional data could give you a competitive advantage • Build a Business-Aligned Data Architecture Optimization Str
ategy
or help sustain it.
• Restore Trust in Your Data Using a Business-Aligned Data Q
3. Establish a data governance framework. uality Management Approach 
• Develop a Master Data Management Strategy and Roadmap
4. Start documenting analytics and AI value chains for your
to Better Monetize Data
organization. They will help you understand how to deliver the “last
mile” and embed AI (and analytics) into decision workflows and
translate insights into actions and impact.

5. Start conversations about an AI-governance framework in the

context of your organization. It will take time to fully develop it, so in
the meantime use this as an opportunity to educate the organization on
AI and build consensus.

Info-Tech Research Group | 99

Info-Tech AI consulting advisory services
Info-Tech’s consulting advisory service provides an impartial, research-based assessment of
your preparedness to proceed with AI, and a roadmap to realize long-term differentiating
value.

Strategy and Governance Foundational Data Management Review

• AI value discovery and roadmap development. • Data profiling, data lineage review, and recommendations.
• AI use case mapping to enterprise value creation (revenue, • Test data management strategy and operationalization.
operating margin, asset efficiency, expectation management). • Master and reference data review and assessment.
• AI accountability and authority change management.

Enterprise Architecture Diagnosis Infrastructure Assessment

• Business process/systems/tooling synthesis. • Big data architecture readiness review.
• Enterprise application analytical content workflow and • Buy vs. build vs. open source AI solution assessment.
efficiency review. • Business operations integration impact assessment.
• Use case coordination with business processes, systems, and
tools.

Info-Tech Research Group | 100

Use the effort/value matrix to support you in prioritizing and sequencing AI pilot projects

3.2 Duration: 30 minutes

Benefit
Overview High
If you are like many other organizations, the resources you have available for
innovation projects are fairly limited. Moreover, with new technologies like AI, it’s best
to start small to fully understand these technologies and their potential, the pre- AI
requisites and dependencies, and to deliver value quickly (“fail fast” approach). project 1
Prioritizing and sequencing pilot projects based on feasibility/effort and
benefits/value will help create that immediate value and help the team focus on what AI
matters most. project 5
AI project
Instructions 6
AI
1. Evaluate the different pilot ideas using the matrix to the right. Consider the project 2
constraints, effort, and benefits associated with each project, as well as pre-
requisites on the previous slide. (“Are You Ready for AI?”) Using these Effort
considerations, plot the initiatives on an effort/benefit map. AI project
2. Use the findings of this map to help prioritize and sequence your AI High 7 Low
initiatives. Your focus should be on the projects that are high on value
AI
and relatively quick to deliver (low effort).
project 3
AI project
4
INPUT Materials
• AI use cases/pilot ideas • Whiteboard or flip chart
• Markers
• Sticky notes
OUTPUT
• Prioritized pilot projects Participants Low

• AI Team Example: Initiative Mapping Exercise

Info-Tech Research Group | 101
For competing ideas, have your AI working groups vote to decide which should be prototyped
first
3.3 Duration: 30 minutes

Instructions
Participants
1. As a group, vote on the ideas/AI use cases that have been generated as a result of • AI project lead
Activity 3.2. Provide a single color of dots. People will only be voting “yes.” You
want to keep this exercise positive and focus on the ideas people connect with • AI working group
instead of highlighting the ones that have not (yet) made the cut.
• There is no minimum or maximum number of dots that can be placed beneath Materials
each idea. A participant can use all votes for one idea.
2. Select the top idea (or two) to move forward to feasibility analysis and prototyping. • Whiteboard and markers
• Voting dots (or same color
markers)
• Sticky notes

Idea List:
Remember, prototyping should be quick and
Idea Idea Idea Idea Idea
4 Idea inexpensive; don’t dismiss ideas under the
1 2 3 5
6
assumption that a prototype would require a
considerable amount of resources. If the idea is
highly beneficial, it deserves to be investigated
further.
8 2 5 4 3 10

Info-Tech Research Group | 102

Use Info-Tech’s Idea Reservoir Tool to document and rank your AI ideas/use cases

Use the Idea Reservoir Tool from Info-Tech’s blueprint Kick-Start It – Lead Business Innovation to record and
rank your AI use cases.
• List your problem statements in the “Problem Statement Documenter” tab.
• Organization’s goals for AI can be set in the “Innovation Goals” tab. These should
be adjusted as organizational strategies shift.
• Use the “Idea Log” tab to record AI use cases, their overall impact, and their
alignment to each of the strategic objectives.
• Rank use cases on the “Shortlist Idea
Generator” tab of the tool. Ranking
should be based on the relative
alignment of ideas with your goals
for AI.
• Decide which use cases to prototype based on this ranking. This
decision should be made as a group.
• Ideas should only be removed from the idea reservoir if prototyping
has demonstrated that the idea has no value to the organization. Keep
a record of removed ideas for future reference.

Update your AI idea reservoir to ensure it always reflects the current status of ideas. If your organization’s
goals change, be sure to change their weightings in the tool; this will impact idea ranking. Store your AI
Idea Reservoir Tool, problem statements, and all relevant documents in a shared and easily accessible
location.
Info-Tech Research Group | 103
Step 2: Create an MVBC for Your First AI PoC

PHASE 3

3.1 3.2
Develop Your AI PoC Create an MVBC for Your
Project Ideas First AI PoC

This step will walk you through the following activities: This step involves the following participants:
• Create an MVBC for the selected AI use case. • AI working group.

Outcomes of this step

• A strong MVBC for your first AI project has been created and submitted for approval by the Executive Sponsor and the
governance bodies.

Info-Tech Research Group | 104

What’s an MVBC?

An MVBC will help you improve speed to value.

Like a minimum viable product (MVP) ‒ a product with just enough features to satisfy early customers and
provide feedback for future product development ‒ a minimum viable business case (MVBC) is a light-weight
document containing enough information to make a go/no-go decision on innovation projects such as those
exploring the potential of AI.

Focusing on the value proposition and the

expected benefits and leaving out the technical
details (for the prototyping phase) will help
you avoid analysis paralysis and get started
sooner on exploring the idea and its potential
to deliver value to your organization.

Image source: Wikipedia, “Motorwagen Serienversion,” n.d. Public Commons.

Info-Tech Research Group | 105

Create an MVBC for your first AI PoC

A full business case is often a poor fit for innovative

projects such as AI, and it poses the risk of wasting
Download Info-Tech’s
momentum as stakeholders wait for paperwork to be done,
Minimum Viable Business Case Documen
causing their interest to begin to wane. t
Use Info-Tech’s .
Fund Innovation With a Minimum Viable Business Case
blueprint to create an MVBC for your first AI PoC project.

Info-Tech’s MVBC Methodology

1 Create a Solution to Someone’s

Problem

2 Reduce the Impact of Your Project

3 Build the Case and Secure Approval

4 Implement, Measure, and Scale

Info-Tech Research Group | 106

Use Info-Tech’s Prototyping Workbook to flesh out the details of your
solution to help build a strong MVBC
As you are completing the MVBC for your first AI PoC, you might want to start thinking
about what the solution could look like and document this in the Prototyping Workbook.
For more details on how to complete the Prototyping Workbook, consult Info-Tech’s
blueprint Kick-Start It-Lead Business Innovation.

"Examine automaton from a customer experience standpoint, a business

standpoint, an application standpoint, and an infrastructure standpoint."
‒ Shankar Balasubramanian, HEXAWARE, in EY, n.d.

"Consider human-to-human, human-to-machine, but also

machine-to-machine interaction." 
‒ Cindy Gallagher, CEO, Liberty Source Edgeverve, in AIIA Network, 2019

Info-Tech Insight
You will be able to use the information documented in the Prototyping Workbook to
make a strong case for the PoC and ultimately for investing in the solution you are
proposing.
Info-Tech Research Group | 107
Beyond your first AI PoC: Connecting your data
systems should be your utmost priority
Why? Because…
"As a rule of thumb, you can expect the transition 
of your enterprise company to machine learning will "10% of the work is AI, 90% of the work is
be about 100x harder than your transition to actually data extraction, cleansing, normalizing,
mobile." wrangling."
‒ Peter Skomoroch, CEO of SkipFlag, in “This Is Why AI Has Yet to Reshape  ‒ Sanjay Srivastava, CDO, Genpact, in “This Is Why AI Has Yet to
Most Businesses,” MIT Technology Review, 2019 Reshape Most Businesses,” MIT Technology Review, 2019

If your organization, like many other organizations, has been collecting data from your operational and transactional systems into separate,
disconnected repositories and is just starting to build the bridges between these silos, expect to spend some time and effort integrating your data.
You can still start with AI, but you will only be scratching the surface. You are not alone though – “the most common uses of AI so far have involved
business processes that are siloed but nonetheless have abundant data, such as computer security or fraud detection at banks” (MIT Technology
Review, “This Is Why AI Has Yet to Reshape Most Businesses,” 2019).

To achieve transformational impact, you will need to link your organization’s data
sources and systems in order to:
1) Train on AI – remember, it thrives on large quantities of quality data.
2) Deliver results to the end-users in an intuitive, seamless manner.

To achieve this, you will need to put your best people on the task of data
Info-Tech Research Group | 108
integration ASAP.
Summary of
Accomplishment
Knowledge Gained Deliverables Completed
• A solid understanding of what AI really is in the modern world.
• Understanding of technologies comprising AI and the difference between general and AAI.  AI mandate and charter are
drafted, and the team involved in
• A clear picture of the types of problems that AI can and cannot help to solve.
this initiative is identified.
• A good understanding of how AI is impacting business functions.
 Applicable use cases are identified
• Understanding of how your peers are using AI and where AI is delivering value in your industry and contextualized.
and other verticals.
 AI ideas are developed and
Process Optimized socialized.
• Defining strategic objectives of your AI.  Problem statements are
• Determining the business goals to get value from AI in your organization. documented for selected ideas.
• Identification of use cases, business problems, and opportunities that AI could  Selected AI ideas are prioritized
potentially help solve within your organization. and sequenced.
• Brainstorming your AI PoC projects.
 First AI PoC is selected, and
• Creating a short list of AI projects. MVBC is created and submitted
• Identifying prerequisites and dependencies and determining feasibility. for approval.
• Selecting your first AI PoC.
• Creating an MVBC for your first AI PoC. Info-Tech Research Group | 109
Summary of If you would like additional
Accomplishment support, have our analysts
guide you through other
Deliverables Completed phases as part of an Info-
 AI mandate and charter are drafted, and the team involved in this initiative
Tech workshop.
is identified.
 Applicable use cases are identified and contextualized. Contact your account representative for
 AI ideas are developed and socialized. more information.
 Problem statements are documented for selected ideas. [email protected]
 Selected AI ideas are prioritized and sequenced.
1-888-670-8889

 First AI PoC is selected, and MVBC is created and submitted for approval.

Info-Tech Research Group | 110

Related
Info-Tech Listed below are additional Info-Tech resources that might be of interest to you.
(This is not an exhaustive list.) To access this research, click on the links below.
Research
• CIO Trend Report 2019 • Insurers Should Use Artificial Intelligen
ce to Thrive in the Age of Disruption
• Activate Your Augmented Reality Initiative
• FIBO: Financial Industry Business Ont
• Artificial Intelligence in Education: Great V ology
alue, Great Challenges Ahead
• Social Ethics of Big Data Mining and AI
• Strategies for Enabling an AI Future in the in Banking
Public Sector
• Design, Ethics, and AI in the Automotive
• Hit The Ground(water) Learning, Not Runn Industry
ing
• Emotion AI Critical to the Future of Aut
• The Road to ITSM AI: Develop an ITSM AI onomous Vehicles
Strategy
• Rise of the Machines: Machine Learnin
g and Enterprise IT Infrastructure

Info-Tech Research Group | 111

Bibliography
Agrawal, Ajay, et al. Prediction Machines: The Simple Economics of Bhatt, Shweta. “5 Things You Need to Know about Reinforcement
Artificial Intelligence. Harvard Business Review Press, 2018. Print. Learning.” KDnuggets. Mar. 2018. Web.

AIIA Network. “AI Live Americas Wrap-Up Report: Distilled Takeaways BPM Partners. “The Importance of Artificial Intelligence in Corporate
From Each Session.” AIIA Network. 2019. Web. Budgeting, Planning and Forecasting.” BPM Partners. Dec. 2018. Web.

Atomwise. “New Ebola Treatment Using Artificial Intelligence: Case Study.” Bradley, Philippa. “Viewpoint: Artificial Intelligence and Local
Atomwise. 24 Mar. 2015. Web. Government.” LGiU. 21 Feb. 2018. Web.

Automated Insights. “Natural Language Generation: New Automation and Bughin, Jacques, and Eric Hazan. “Five Management Strategies for Getting
Personalization Opportunities for Business.” LinkedIn SlideShare. 26 Jan. the Most From AI.” MIT Sloan Management Review. 19 Sep. 2017. Web.
2016. Web.
Cann, Geoffrey. “The Three Ways AI Will Impact Oil and Gas.” EnergyNow
Bergstein, Brian. “This Is Why AI Has Yet to Reshape Most Businesses.” MIT Media. 6 Aug. 2018. Web.
Technology Review. 13 Feb. 2019. Web.
Capgemini Consulting. “Turning AI Into Concrete Value: The Successful
Berman, Daniel. “The Growing Role of Machine Learning in Monitoring.” Implementers’ Toolkit.” Capgemini Consulting. 2017. Web.
Logz.io. 29 May 2017. Web.
Carnegie Learning. “MATHiaU Is Here to Solve Your Developmental Math
Berruti, Federico, et al. “Intelligent Process Automation: The Engine at the Problem.” Carnegie Learning. n.d. Web.
Core of the Next-Generation Operating Model.” McKinsey & Company. Mar.
Catalysts. “Computer Vision for Quality Assurance.” Catalysts. n.d. Web.
2017. Web.

Bersin, Josh. “AI in HR: A Real Killer App.” Forbes. 18 Jun. 2018. Web.

Info-Tech Research Group | 112

Bibliography
Celaton. “Case Study Improving Customer Service With Machine Learning, Derandomized. “Khan Academy: Machine Learning → Measurable
Virgin Trains.” Celaton. n.d. Web. Learning.” Derandomized. 30 May 2019. Web.

Chen, Stephen. “China’s Schools Are Quietly Using AI to Mark Students’ Digital Health. “Babylon Health to Power NHS 111 With ‘AI Triage’ Bot.”
Essays...But Do the Robots Make the Grade?” South China Morning Post. Digital Health. Jan. 2017. Web.
28 May 2018. Web.
Dolgin, Elie. “AI Face-Scanning App Spots Signs of Rare Genetic
Cloud AutoML. “Google Cloud.” Cloud AutoML. n.d. Web. Disorders.” Nature. 7 Jan. 2019. Web.

Collier, Matt, et. al. “‘Artificial Intelligence: Healthcare’s New Nervous Domingos, Pedro. The Master Algorithm: How the Quest for the Ultimate
System.” Accenture. 2017. Web. Learning Machine Will Remake Our World. Basic Books, 2015. Print.

CTI. “The Notion of Artificial Intelligence Is a Thing of The Past. The Drozdov, Sam. “An Intro to Machine Learning for Designers.” UX
Reality of Artificial Intelligence Is a Thing of Today.” CTI Content Collective. 30 May 2018. Web.
Technologies Inc. n.d. Web.
Eckerson, Wayne W. “The Impact of AI on Analytics: Machine-Generated
Daily Mail. “The 3D Printed ‘Digital Twin’ of Your Heart That Could Save Intelligence.” Eckerson Group. 1 Oct. 2018. Web.
Your Life: AI System Lets Surgeons Simulate Surgery.” Daily Mail. 31 Aug.
Empatica Support. “Event Detection Accuracy.” Empatica Support. 7 Feb.
2018. Web.
2019. Web.
Darktrace. “Fight Back in Real Time.” Darktrace. n.d. Web.
Ereth, Julian, and Wayne Eckerson. “AI: The New BI How Algorithms Are
Daugherty, Paul, and H. James Wilson. Human + Machine: Reimagining Transforming Business Intelligence and Analytics.” Eckerson Group. Oct.
Work in the Age of AI. Harvard Business Review Press, 2018. Print. 2018. Web.

Davenport, Thomas H. The AI Advantage: How to Put the Artificial Info-Tech Research Group | 113
Intelligence Revolution to Work. The MIT Press, 2019.
Bibliography
Ereth, Julian, and Wayne Eckerson. “DataOps: Industrializing Data and GuildOne. “GuildOne’s Royalty Ledger Settles First Royalty Contract on
Analytics.” Eckerson Group. 30 Jun. 2018. Web. R3’s Corda Blockchain Platform.” GuildOne. 14 Feb. 2018. Web.

EVRY. “The New Wave of Artificial Intelligence.” Whitepaper. EVRY. Glenn, Stacia. “Burglaries Dropping in Tacoma.” The News Tribune. 18
2017. Web. Oct. 2015. Web.

EY. “Intelligent Automation: Reshaping the Future of Work With Robots.” Gluck, Josh. “AI and Healthcare: Cure-All, Poison Pill, or Simply Smarter
EY. n.d. Web. Medicine?” Hea!thcare Innovation. 9 Jan. 2019. Web.

FDA News. “FDA Permits Marketing of Artificial Intelligence-Based Goldblatt, Ran, et al. “Artificial Intelligence for Smart Cities: Insights From
Device to Detect Certain Diabetes-Related Eye Problems.” FDA News Ho Chi Minh City’s Spatial Development.” The World Bank Blogs. 8 Mar.
Release. US Food & Drug Association. 11 Apr. 2018. Web. 2018. Web.

Fingas, Jon. “Speedy AI Image Analysis Could Help Doctors During Gonfalonieri, Alexandre. “What I Have Learned After Several AI Projects.”
Surgery.” Engadget. 19 Jun. 2018. Web. Medium. 12 Jan. 2017. Web.

Galeon, Dom. “An AI Completed 360,000 Hours of Finance Work in Just Gopal, Vikas. “Transform Cost Optimization With RPA and Machine
Seconds.” Futurism. 8 Mar. 2017. Web. Learning.” Tata Consultancy Services. 2017. Web.

Garfield, Michael. “The Future Is Indistinguishable From Magic.” Medium. Government of Canada. “About GPHIN.” Government of Canada. n.d.
2018. Web. Web.

Gawande, Atul. “The Checklist: If Something So Simple Can Transform Grothaus, Michael. “An AI Can Now Write Its Own Code.” FastCompany.
Intensive Care, What Else Can It Do?” The New Yorker. 10 Dec. 2007. 4 Apr. 2018. Web.
Web.
Info-Tech Research Group | 114
Bibliography
Hadsell, Raia. “Artificial Intelligence, Video Games and the Mysteries of Kalis, Brian, et al. “10 Promising AI Applications in Health Care.” Harvard
the Mind.” YouTube, uploaded by TEDx Talks. 28 Nov. 2017. Web. Business Review. 10 May 2018. Web.

Han, Meghan. “Six Disruptive AI Platforms for Law.” Synced Review, A Knight, Will. “Cheaper A.I. for Everyone Is the Promise With Intel and
Medium Corporation. 9 Oct. 2017. Web. Facebook’s New Chip.” MIT Technology Review. 14 Jan. 2019. Web.

Harvard Business Review. “Meet Your New Employee: How to Manage the Knockri. “Main Webpage.” Knockri. n.d. Web.
Man-Machine Collaboration.” Harvard Business Review. Jun. 2015. Web.
Kozyrkov, Cassie. “Top 10 Roles in AI and Data Science.” HackerNoon. 26
Hasenfratz, Linda. “AI at Linamar.” YouTube, uploaded by Creative Jul. 2018. Web.
Destruction Lab. 1 Nov. 2018. Web.
Lamarre, Eric, and Brett May. “Ten Trends Shaping the Internet of Things
Hortonworks. “Data Science With Hadoop.” LinkedIn SlideShare. Fall Business Landscape.” McKinsey & Company. Jan. 2019. Web.
2014. Web.
LaMotta, Lisa. “Pharma and AI? Let’s Try Augmented Intelligence First.”
IBM. “Turning Data Into Chemicals and Petroleum Insights: How the Biopharma Dive. 23 Jul. 2018. Web.
Industry Is Becoming Cognitive.” IBM. May 2017. Web.
Larisa. “What Do You Mean by Machine Learning?” TechJini. 16 Feb.
Infosys. “Amplifying Human Potential: Towards Purposeful Artificial 2017. Web.
Intelligence.” Infosys. 2017. Web.
Library of Congress. “Wright Brothers.” Library of Congress. n.d. Web.
J.H. “The Robo-Tutor Is Coming.” The Economist. 13 Jul. 2018. Web.
Light, Nanette. “Virtual World May Help Those With Autism Sharpen
Kahn, Jeremy. “The AI That Spots a Stopped Heart.” Bloomberg. 20 Jun. Social Skills.” The Dallas Morning News. 9 Jul. 2018. Web.
2018. Web.
Info-Tech Research Group | 115
Bibliography
Loucks, Jeff, et al. “State of AI in the Enterprise, 2nd Edition - Early McKay, Stephanie, and Sulzhan Bali. “Disruptive Technologies: Keeping an
Adopters Combine Bullish Enthusiasm With Strategic Investments.” AI on Disease Surveillance.” Global Health Now. John Hopkins Bloomberg
Deloitte Insights. 22 Oct. 2018. Web. School of Public Health. 25 Apr. 2018. Web.
McKinsey & Company. “Notes From the AI Frontier: AI Adoption
Manyika, James, and Jacques Bughin. “The Promise and Challenge of the
Advances, But Foundational Barriers Remain.” McKinsey & Company.
Age of Artificial Intelligence.” Executive Briefing. McKinsey & Company.
Nov. 2018. Web.
Oct. 2018. Web.
McKinsey & Company. “The Executives AI Playbook.” McKinsey &
Marr, Bernard. “The Amazing Ways Verizon Uses AI And Machine Company. 2018. Web.
Learning To Improve Performance.” Forbes. 22 Jun. 2018. Web.
McNeill, Sam. “Artificial Intelligence in the Classroom.” Microsoft. 1 Mar.
Martin, Sean. “AI Warning: Machines Will Replace Humans at the Top – 2018. Web.
And Wipe Us Out if We Dare Resist.” Express. 18 Sep. 2019. Web.
Microsoft. “Seeing AI.” Microsoft. n.d. Web.
Matei, Adrienne. “Robots and the Future of Dying: How Artificial
National Science Foundation. “Fighting Food Poisoning in Las Vegas With
Intelligence Is Changing End of Life Care.” NUVO. 26 Jul. 2018. Web. Machine Learning.” National Science Foundation. 7 Mar. 2016. Web.
McFadden, Christopher. “These 7 AI-Powered Doctor Phone Apps Could Neuromation. “The Role of Artificial Intelligence in Education.”
Be the Future of Healthcare.” Interesting Engineering. 24 Jan. 2019. Web. Medium.10 Jun. 2018. Web.
McGraine, Clare. “Amazon Unveils New Service to Mine and Decode Novatio. “10 Common Applications of Artificial Intelligence in
Medical Records Using Artificial Intelligence.” GeekWire. 27 Nov. 2018. Healthcare.” Novatio. n.d. Web.
Web.
Office of the CTO Blog. “AI at VMware: A History, Present and What’s
McGregor, Caitlin. “AI Spotlight: Future of Work and Talent AI With Next.” OCTO Blog. 2 Nov. 2017. Web.
Caitlin McGregor of Plum.” Elevate 2018 Conference. 21-27 Sept. 2018, Info-Tech Research Group | 116

Toronto, ON.
Bibliography
Plum. “Human Potential Quantified Like Never Before.” Plum. n.d. Web. SAS Blogs. “Why People and Process Matter, in Addition to Great
Technology, in Predictive Analytics.” SAS Blogs. 11 Jan. 2013. Web.
Pontin, Jason. “Greedy, Brittle, Opaque, and Shallow: The Downsides to
Deep Learning”. Wired. 2 Feb. 2018. Web. Schlecht, Daniel, et al. “Artificial Intelligence: The Ultimate Disrupting
PwC. “Bot.Me: A Revolutionary Partnership - How AI Is Pushing Man and Force for Oil and Gas Companies.” The First SPE Norway Magazine. Sep.
Machine Closer Together.” Consumer Intelligence Series: PwC. 2017. Web. 2017. Web.

PwC. “What Doctor? Why AI and Robotics Will Define New Health.” PwC. Science Serve. “Helping People Walk With Exoskeletons | The Growth of
Jun. 2017. Web. Wearable Robots in Healthcare.” Science Serve: Dr. Hempel Digital Health
Network. 9 Jul. 2018. Web.
Ransbotham, Sam, et al. “Artificial Intelligence in Business Gets Real:
Pioneering Companies Aim for AI at Scale." MIT Sloan Management Sennaar, Kumba. “Artificial Intelligence in Oil and Gas – Comparing the
Review. 17 Sep. 2018. Web. Applications of 5 Oil Giants.” Emerj. 18 Feb. 2019. Web.

Ransbotham, Sam, et al. “Reshaping Business With Artificial Intelligence: Sivers, Derek. “First Follower: Leadership Lessons from a Dancing Guy.”
Closing the Gap Between Ambition and Action.” MIT Sloan Management Derek Sivers. 11 Feb. 2011. Web.
Review. Fall 2017. Web.
Smith, Steven J. “The Rise of Data Science Platforms: Key Features for
Sandewall, Erik. “The Goals of Artificial Intelligence Research: A Brief Automating Analytics and Driving Value.” Eckerson Group Research. Nov.
Introduction.” Knowledge Representation Framework Project. 8 Aug. 2010. 2018 Web.
Web.
Sommer, James. “Is AI Riding a One-Trick Pony?” MIT Technology Review.
Sano, Michael. “Can AR/VR Improve Learning? Integrating Extended 29 Sep. 2017. Web.
Reality Into Academic Programs #DLNchat.” EdSurge. 19 Jul. 2018. Web. Info-Tech Research Group | 117
Bibliography
Somers, James. “The Man Who Would Teach Machines to Think.” The The Wall Street Journal. “The AI Doctor Will See You Now.” The Wall
Atlantic. Nov. 2013. Web. Street Journal. n.d. Web.

Stitch Fix. “Algorithms Tour: How Data Science Is Woven Into the Fabric Turner, Karen. “This Robot Lawyer Helps the Newly Evicted File for
of Stitch Fix.” MultiThreaded. n.d. Web. Housing Aid.” The Washington Post. 9 Aug. 2016. Web.

Swift Medical. “Better Outcomes For All.” SWIFT Medical. n.d. Web. University Hospitals. “Proactive Staffing and Patient Prioritization to
Decompress ED and Reduce Length of Stay.” University Hospitals. 2018.
TeachThought Staff. “10 Roles For Artificial Intelligence In Education.” Web.
TeachThought. 16 Sep. 2018. Web.
US Citizenship and Immigration Services. “Meet Emma, Our Virtual
Tech Advantages. “BMW Factory Humans & Robots Work Together At Assistant.” US Citizenship and Immigration Services. n.d. Web.
Dingolfing Plant.” YouTube, uploaded by BMW GroupMedia. 2017. Web.
Vazquez, Favio. “Deep Cognition.” KDnuggets. 2017. Web.
Tech Jini. “What Do You Mean By Machine Learning.” Tech Jini. 16 Feb.
2017. Web. Verdict Medical Devices. “AI-Based Blood Test to Identify Cancer at the
Earliest Stages.” Verdict Medical Devices. 16 Nov. 2018. Web.
The Economist. “Artificial Intelligence: Anything You Can Do, AI Can Do
Better.” The Economist. n.d. Web. von Radowitz, John. “Intelligent Machines Will Replace Teachers Within
10 Years, Leading Public School Headteacher Predicts.” Independent. 11
The Guardian. “Computer Learns to Detect Skin Cancer More Accurately Sep. 2017. Web.
Than Doctors.” The Guardian. 29 May 2018. Web.
Walker, Jasson Jr. “4 Government Agencies Using Bots.” Medium. 21 Sep.
2017. Web.
Info-Tech Research Group | 118
Bibliography
Walker, Jasson Jr. “Azure Government Chatbot Reduces Email By 50%.” Wolfram Research. “Building a Data Scientist Pipeline.” Wolfram Research.
cFocus. 5 May 2017. Web. 17 Aug. 2017. Web.

Wikipedia. “Artificial Intelligence.” Wikipedia. n.d. Web. World Intellectual Property Organization. “Index of AI initiatives in IP
Offices.” WIPO. n.d. Web.
Wikipedia. “Automated Species Identification.” Wikipedia. n.d. Web.
World Oil. “Digital Transformation Journey in Upstream Oil and Gas, Gas
Wikipedia. “Deep Learning.” Wikipedia. n.d. Web. Processing.” World Oil. 6 Sep. 2018. Web.
Wikipedia. “Explainable Artificial Intelligence.” Wikipedia. n.d. Web. Yi, Shu Ng. “China Is Using AI to Predict Who Will Commit Crime Next.”
Mashable. 24 Jul. 2017. Web.
Wikipedia. “Human-Computer Interaction.” Wikipedia. n.d. Web.
Yonah. “VisionHack 2017: The State of Computer Vision in Russia.” WM.
Wikipedia. “Motorwagen Serienversion.” Wikipedia. n.d. Web.
14 Aug. 2019. Web.
Wikipedia. “Neuron3.” Wikipedia. n.d. Web.
Zahid, Zahra. “Practice Area Lead – Modern Data Architecture and AI.”
Wikipedia. “People Counter.” Wikipedia. n.d. Web. Slalom Consulting. 2018. Interview.

Wikipedia. “Transfer Learning.” Wikipedia. n.d. Web. Zunger, Yonatan. “Asking the Right Questions About AI.” Medium.11 Oct.
2017. Web.
WildFire. “AI Use Cases in Oil and Gas Industry.” YouTube, uploaded by
WildFire. 13 Mar. 2018. Web.

Info-Tech Research Group | 119

Info-Tech Research Group | 120