Taking AI to the next level

Harnessing the full potential and value

of AI while managing its unique risks

Artificial intelligence (AI) has become a clear and undeniable force for business
disruption and transformation. It is delivering significant business benefits today—and
its potential to shape the future is even greater. In fact, the main challenge now is how to
scale up AI-led innovations to deliver the greatest impact as efficiently and effectively as
possible while at the same time managing AI’s unique risks.

Although AI value and AI risk management are often viewed The answers to these questions will be different for every company;
as separate and distinct, these four key questions can help however, the questions themselves are universally applicable. Also,
you address both factors simultaneously: there are no perfect answers. What matters most is being intentional
and considering each as you navigate your AI journey and avoid
1. Value: How do you predict, prioritize and capture
falling too far behind.
the value of AI for your business?

2. Architecture: What architectural capabilities are The AI systems market is expected to reach $79 billion in 2022.1 Yet a
needed to scale AI? recent Deloitte survey found that while 79 percent of companies are
already experimenting with AI, less than one percent are using AI
3. Workforce: How will AI affect your workforce?
widely across the enterprise.2
4. Governance: How do you tackle the challenges of AI
responsibility, ethics and governance to harness AI’s full In this report, we explore getting beyond experimentation through
potential and value? a detailed look at each of the four key questions raised above. But
before going into the details, let’s briefly cover the basics. What is AI?
How is it different? And why should you care?
Taking AI to the next level | Harnessing the full potential and value of AI while managing its unique risks

Artificial intelligence demystified

Theorists have been arguing for decades about what constitutes AI’s “intelligence” is based on underlying
true “intelligence”—both in machines and in people—and that capabilities similar to those of human intelligence
debate might never end. However, in practical terms, a good working The core capabilities of AI are similar to those of human intelligence,
definition of AI is: Computer systems able to perform tasks normally such as clustering (pattern recognition), categorization, anomaly
requiring human intelligence. detection and regression and prediction. However, AI can effectively
apply these capabilities to data sets and challenges that are far
Key things to know about AI richer and more complex than humans can readily handle.
AI systems learn
Unlike traditional computer systems that are explicitly programmed There are many ways to achieve AI
to follow sets of rules and produce deterministic outcomes, AI AI systems use a variety of methods and algorithms, including
systems get smarter over time, sometimes on their own. This ability symbolic reasoning, Bayesian inference, connectionist AI (neural
to learn gives AI the potential to deliver superior outcomes—and networks), genetic algorithms that emulate evolution, and learning
to solve problems that previously were unsolvable or poorly solved by analogy. They also use a variety of learning approaches, including
(e.g., using computer vision to analyze medical scans, or using voice supervised, unsupervised and reinforcement. As such, AI is not just
recognition to enable virtual assistants such as Siri, Alexa, Cortana, a single technology; it’s a rich set of development techniques and
and Google Home). However, AI’s learning ability also produces problem-solving approaches. The better you understand AI, the
outcomes that can be challenging to explain and are at times more impactful your company’s AI-driven innovations can be.
probabilistic (i.e., subject to chance and variation), which contributes
to concerns regarding AI trustworthiness. In today’s marketplace, AI-based capabilities are emerging that are
proving to be valuable (Figure 1).

Figure 1. A sampling of AI capabilities

Natural
Virtual Sentiment
language
agents analysis
processing

Content Image
extraction recognition
Recommendation

AI-related innovations like these are reshaping the business landscape and propelling forward-thinking companies into the future—enabling
them to do different things and do things differently (even rethinking their business and operating models).

For example, Ant Financial uses AI and data from its mobile-payments platform (AliPay) to serve 10 times as many customers as the largest
US banks (more than a billion customers) with one-tenth as many employees. The company is built around a data-centric business model and
digital core, generating unique and valuable innovations and insights without the operating constraints that burden traditional firms.3

2
 Taking AI to the next level | Harnessing the full potential and value of AI while managing its unique risks

How do you predict, prioritize

and capture the value of AI?

Although AI’s overall ability to create business value is now widely Here’s a structured model for thinking about existing and future
accepted, the specific impacts AI is likely to have are not as well ways that AI can create business value. It’s made up of five
understood and vary by company and industry. fundamental levers (Figure 2):

• Intelligent automation. Automate the “last mile” of automation

At the industry level, how is your industry likely to be affected
by freeing humans from low-value and often repetitive activities
given AI’s continued maturity? Will the impact be disruptive or
(which, ironically, are in many cases conducted to gather and
evolutionary? Will your industry be destroyed (the way digital
prepare data for machines).
technology destroyed travel agencies and film photography) or
simply transformed (the way the music industry has shifted from • Cognitive insights. Improve understanding and decision-making
records and CDs to paid downloads and streaming services)? Will through analytics that are more proactive, predictive, and able
industry boundaries blur through convergence with other new or to uncover patterns in increasingly complex sources.
existing industries?
• Transformed engagement. Change the way people interact with
technology—allowing machines to engage on human terms
At the company level, how can AI enable your business to create
rather than forcing humans to engage on machine terms.
and capture shifting value pools, create sustained competitive
advantage, and achieve profitable growth? How will AI affect your • Fueled innovation. Redefine “where to play?” and “how to win?” by
business strategy? How can AI help you achieve your strategic goals? enabling creation of new products, markets and business models.
How will it affect various business functions and how they work in
• Fortified trust. Secure the franchise from risks such as fraud
the future? What problems do you want/need to solve with AI? What
and cyber, improve quality and consistency, and enable greater
quantifiable value will you get from investing in AI? How should you
transparency to enhance brand trust.
prioritize your AI initiatives and develop a plan to execute?

Unconventional Wisdom – Part 1

Don’t judge AI by its past, or even its present.

Judge AI by its future (which is right around the corner).
In the earliest stage of the AI S-curve, most companies focused on using intelligent automation to reduce costs associated with
repetitive, labor-intensive processes. However, AI technology is maturing rapidly and we are now at the stage where opportunities
for disruption and revenue generation are coming to the fore. When thinking about how to apply AI to your business going
forward, be sure to consider all five value levers—not just cost reduction through automation.

3
Taking AI to the next level | Harnessing the full potential and value of AI while managing its unique risks

Figure 2. The five levers for AI value

Intelligent Cognitive Transformed Fueled Fortified

automation insights engagement innovation trust
Before AI Before AI Before AI Before AI Before AI
People engage in Insights are limited People engage Innovation is limited Reactive and rule-
repetitive low-value by static and reactive with systems and by current technology oriented security pre-
activities, often at the analytic structures that organizations on paradigms, as well as vention and detection;
service of machines struggle to deal with machine terms the growing complexity transparency in decline
(for example, cleansing, increasingly complex, (for example, filling of rules and data. as organizations and
preparing, and keying in distributed, real-time in forms). information increase
data; handling system information. After AI in complexity.
exceptions). After AI New market, product,
After AI Machines engage with and service opportuni- After AI
AI can identify patterns people on human terms ties emerge as human Dynamic and proactive
After AI through proactive and (for example, natural capabilities scale up security measures are
AI systems “read” predictive analysis of language, gestures) massively with the help a priority, as well as
unstructured data, vast amounts of highly and customers, em- of AI systems that learn. increased transparency
optimize and repair complex information ployees, and other key into decision-making
themselves, and are (that is often in motion). stakeholders receive a (with less human bias).
overseen by people. hyper-personalized and
efficient multichannel
experience.

4
 Taking AI to the next level | Harnessing the full potential and value of AI while managing its unique risks

What architectural capabilities

are needed to scale AI?

AI requires the right supporting architecture and infrastructure. Fundamental elements of an AI-enabling
Without that solid foundation, your AI solutions will be severely technology architecture
limited, prohibitively expensive, or just won’t work. Cloud
Public cloud has become the essential enabler for business
Digital native firms are at the forefront in this area. Microsoft, for innovation and disruptive technologies (including AI). It provides
example, is transforming itself into an AI-driven firm by reorganizing flexible and scalable access to a level of computing resources that
its internal IT and data assets and migrating all legacy processes would be physically or economically impossible to achieve using
onto a consistent software base and data architecture in the traditional in-house IT service models. It also provides fast and easy
Microsoft Azure cloud. The company’s new centralized operating access to valuable data that historically was buried deep inside
platform helps integrate the organization using a shared software organizational silos. Just as important, cloud provides instant access
component library, algorithm repository and data catalog—enabling to the latest AI innovations—and to pre-built, pre-configured building
the company to scale and deploy digital processes across different blocks that can help your company jumpstart its AI innovation
segments of the organization.4 efforts. Cloud is no longer just a means to reduce operating costs; it
is the platform for digital business and is where innovation happens.
Given the rising importance of AI—and digital technology in
general—every business leader needs both a practical and Data
theoretical understanding of the latest technology innovations. The data that AI learns from is just as important as the algorithms
AI is no longer just the domain of the CIO and CTO. Because in an that allow it to learn. In fact, 60 percent of companies surveyed by
increasingly digital world, technology is having an impact on every Forrester cited managing data quality as a key challenge in delivering
aspect of business, which means that in many ways every company AI capabilities.5 Data that is static, narrow and disjointed severely
is now a technology company, and every business leader is a limits the AI solutions that can be built, their learning potential and—
technology leader. consequently—their impact. Modern data architectures that provide
holistic and secure access to quality first-, second- and third- party
data (in motion and at rest) are foundational to innovative AI
development and the operating models of the future.

AI is no longer just the domain of the CIO and CTO. In many

ways every company is now a technology company, and every
business leader is a technology leader.

5
Taking AI to the next level | Harnessing the full potential and value of AI while managing its unique risks

Other considerations to keep in mind

• What do the capabilities and components of an AI-enabled • How do you leverage enterprise architecture to support
architecture look like? the scaled development of proofs of concepts across the
organization? What is your data strategy (data sources
• What technology capabilities, data, and systems are required
and signals, data collection, data preparation, acceptance
to enable your AI priorities?
for AI, models)?
• How does AI impact your existing technical architecture?
• How will you access and control first-, second- and third-
How do you need to change this to enable AI scaling?
party data sources?
• What will you build vs. acquire from hyperscale cloud
providers vs. niche third parties? Why? How will your Companies should consider taking a more active role in AI
AI services interrelate and build upon each other? innovation. Future AI breakthroughs are likely to stem from
close collaboration with and among AI start-ups, technology
• What vendors and ecosystems will you leverage to enable
and professional services firms, and academia. Businesses
your AI priorities?
may be able to gain a competitive advantage by actively
• How will you build and measure value from AI proofs cultivating these relationships, instead of relying solely on their
of concepts? own abilities to scan the technology landscape for whatever
interesting developments happen to arise.

Unconventional Wisdom – Part 2

Don’t let past technology investments limit your future.

A company that has invested heavily in technologies that are not AI-enabling will naturally be reluctant to
move away from them. But in order to stay competitive in a future where AI is a fundamental operating
requirement, there may be no choice. For example, companies that have been pouring money into legacy
systems and other in-house IT capabilities for decades will likely need to phase those out quickly in favor of
public cloud. Letting past decisions limit your ability to fully harness AI in the future is self-defeating.

Ultimately, it’s not about investing in AI; it’s about investing in the future of your business. AI just happens to
be a key to that future.

6
 Taking AI to the next level | Harnessing the full potential and value of AI while managing its unique risks

How will AI affect your workforce?

AI has tremendous implications for a company’s workforce. However, When developing an AI strategy, key workforce questions to
those implications are different than what most people think. In consider include:
particular, there is a common misconception that AI equates to
• What will the business look like when the workforce is a
replacing people with machines. Sometimes this does happen,
mix of humans with machines?
especially for the low-skill, labor-intensive activities that have
been a primary focus for early AI efforts, which revolved around • Which processes will be the top priority targets for human/
automation. However, many of AI’s most compelling use cases machine collaboration? What are the challenges?
involve machines and people working together as a team, with AI
• How will the organizational culture need to change?
augmenting humans—not replacing them—and making their jobs
more enjoyable and rewarding. • What training will be needed to teach people to
collaborate with machines?
AI workforce implications can be divided into three broad categories.
• How will performance measurement and compensation
The first focuses on creating a general workforce that can work
be affected?
effectively with intelligent machines. The second focuses on acquiring
and cultivating the business and technical talent necessary to design,
AI development
build and deploy top-notch AI solutions. The third focuses on
AI technology requires a broad range of specialized roles and
accelerating AI development and deployment through centers of
skill sets that go beyond traditional IT capabilities; roles such
excellence and other mechanisms that provide expertise and support
as algorithm developers, machine learning engineers and data
exactly when needed.
scientists. Gaining access to high-quality talent with those unique
skills can have a major impact on AI project outcomes. Key
General workforce
questions to consider include:
AI will create new business roles and enable and transform existing
roles, such as customer service and sales. Even in an ideal case where • What AI-specific skills are needed?
AI perfectly replicates human behavior, workers will still need to get
• Which skills are best maintained in-house?
comfortable with the idea of collaborating with machines (versus just
using them). And in more realistic scenarios, people will also need to • Which skills are already available in the existing workforce, and
learn how to work around AI’s unique quirks and limitations. which need to be shored up?

• What are the options for acquiring scarce AI talent (third-party

vendors and ecosystem partnerships; training/upskilling; external
contractors and experts)?

As noted in the architecture section, expanded relationships with

academia and other external partners can be valuable sources of
talent, expertise, and innovation.

Many of AI’s most compelling use cases involve machines

and people working together as a team, with AI augmenting
humans—not replacing them.

7
Taking AI to the next level | Harnessing the full potential and value of AI while managing its unique risks

Acceleration
Transforming large segments of the workforce by developing new This has improved the quality and pace of service delivery. It has
skills is time-consuming and can inhibit the pace of innovation. also improved technology standardization; enabled the firm to
That’s why many organizations are turning to temporary operating test AI on its most differentiating opportunities; provided sandbox
constructs such as centers of excellence (CoE). These centers environments for learning and experimentation; and helped
typically focus on education, advice, research and knowledge sharing the firm get a head start on addressing issues such as bias and
that accelerate near-term value creation while building a bridge to explainability. Perhaps most important, it has created a sense of
the future workforce. For example, a large investment bank recently progress and helped build enthusiasm and momentum for skills
created an AI CoE that provides project consulting services, platform changes within the organization.
engineering, applied research and governance services.

Unconventional Wisdom – Part 3

AI is less about machines replacing people and more about

machines collaborating with people.
One of the most common fears about AI is that it will steal jobs and make humans obsolete. However, in reality, many
of AI’s most valuable and compelling use cases revolve around people and machines working together to perform
challenging tasks and jointly evolve their capabilities.

In many situations, AI isn’t taking over people’s jobs—it is helping people be more effective and making their jobs more
satisfying and enjoyable by boosting their performance and reducing tedium. Also, it can free people up to spend more
time and energy on activities that are more valuable and strategic to the business.

8
 Taking AI to the next level | Harnessing the full potential and value of AI while managing its unique risks

How do you tackle the challenges of AI

responsibility, ethics and governance?

Because AI often learns and behaves like humans, it presents 3. Privacy. AI systems need to comply with current privacy
many of the same kinds of risks that a company faces in its human standards, including the right to safety, to be informed, to
workforce. As such, without effective safeguards, AI might act in choose, to be heard and to be forgotten.6
ways that are unpredictable or have unintended consequences.
4. Safe/secure. AI systems need to be adequately protected
(See Unconventional Wisdom – Part 4.)
from cyber threats.

Achieving AI that is trustworthy requires effective governance in 5. Responsible/accountable. Clear policies must be in place to
every phase of the AI life cycle. Waiting until after problems arise can determine who is responsible/accountable for AI outputs
have significant legal, regulatory and financial repercussions. It can and actions.
also cause significant reputation damage and embarrassment for a
6. Transparent/explainable. Companies must be willing and
company and its leaders.
able to explain how data is being used and how AI systems
make decisions.
Businesses can and should innovate at a rapid pace; however, they
also need to establish guardrails for responsibility and ethics that
BMO Financial Group is working towards accelerated business
allow AI innovation to proceed rapidly while reducing risk. Even in
outcomes using AI while preserving trust with their customers
industries such as banking that are closely regulated and therefore
and internal stakeholders with their Trustworthy AI initiative. It is
have strong governance in place, enhanced governance will likely
achieving these outcomes by designing for fairness, transparency
be needed to address AI’s unique challenges and risks. Also, AI can
and explainability as initial priorities; rapidly identifying and
affect the controls a business has already established, as well as its
mitigating issues posed by the incremental risks of AI; and
process and life cycle for implementing future controls.
enhancing its development, enterprise risk, privacy, legal and
ethics practices to embrace AI at scale.
Six key attributes of trustworthy AI:

1. Fair/unbiased. AI systems need to include internal/external

checks to ensure equitable treatment for all participants.

2. Robust/reliable. AI systems that learn from humans (and from

other systems) need to produce outputs that are reliable and
consistent with human norms of behavior.

Unconventional Wisdom – Part 4

The good news is AI acts human and thinks for itself.

The bad news is AI acts human and thinks for itself.
Like the humans it seeks to emulate, AI can make mistakes. It can exercise poor judgment and exhibit biases that might seem
perfectly logical based on objective data but that violate social norms. And it can learn bad habits. (The most famous example
being a chatbot that started using offensive language it had learned from human users.)

From a risk perspective, AI’s biggest problems stem from its biggest advantages—reach, scale and speed. Machines and humans
are both imperfect. However, problems created by humans tend to be limited to a relatively small part of the business that is
directly within their reach, whereas problems created by AI can propagate at digital speed across the entire enterprise.

9
Taking AI to the next level | Harnessing the full potential and value of AI while managing its unique risks

The AI journey:
How do we get from here to there?

In the future, AI mastery will likely be essential for businesses in Scaling

every industry. As with the rise of digital, AI will progress from Scaling is often characterized by moving beyond narrow classes
something companies need to do toward something they need to of use cases and subsets of business functions to a broad
be. Today, many organizations are at least somewhat satisfied with portfolio of valuable AI efforts; fully embracing public cloud for
their current AI efforts because they are doing a lot of different AI services; experimenting with, developing and executing a
things that can be considered AI. However, they have a sense that metadata management framework for first-, second- and third-
more can and should be done. In particular, they need to bridge party data; and institutionalizing the AI development life cycle.
the gap from experimenting and doing to being an organization Additionally, organizations in the scaling stage are often formalizing
with AI as a core capability. As with most disruptions, the AI journey efforts to raise awareness of AI’s impact; upskilling and reskilling
involves three major stages: establishing a foundation, where a their workforce; embracing non-traditional partnerships; and
company begins to gain confidence in AI’s potential and its ability establishing policy frameworks, accountability, playbooks and
to harness it; scaling, where the company can create AI at a rapid monitoring for AI governance.
pace through mechanisms that are well governed and not limited
by function or category; and mastery, where the company’s brand, Mastery
business and operating model are shaped by AI. Every company’s Organizations that have achieved mastery are typically moving
journey through the three stages will be unique, but all will require beyond a use-case and function-based portfolio to transforming the
maturation across the four dimensions of value, architecture, enterprise and driving widespread innovation of products, services,
workforce, and governance. and markets. They are adopting a data-first, cloud-powered,
hypothesis generating and testing business and technology
Establishing a foundation architecture. Most are having success retaining and attracting the
Organizations at this stage of maturity are typically agreeing on best next-generation talent. Creating AI that benefits communities
a definition of AI, why it’s different and why it matters. They are and society is common. And integrating AI governance into corporate
clearly articulating the sources of value the organization will pursue and technology governance, establishing trust as a competitive
over time, as well as establishing near-term priorities. Most will be differentiator, is the norm. Aside from some of the world’s leading
embracing public cloud for targeted use cases and beginning to digital companies, most businesses are still in the early stages
industrialize the data pipeline. Establishing a CoE to broker learnings of the AI journey. Now is the time for decisive action. Companies
and highlight skill availability is common at this stage, as is identifying today need to start or accelerate their journey in an intentional and
and agreeing to guiding principles and an overarching governance structured way that anticipates their competitors feverishly racing to
framework for AI (centralized, distributed or hybrid). achieve AI mastery.

Unconventional Wisdom – Part 5

AI isn’t something companies need to do. It’s something they need to be.
At the moment, most companies still think of AI in terms of isolated use cases and pilot programs. However, in order to harness
the full potential and value of AI, it must be deeply ingrained in every part of the enterprise, from strategy and technology to
workforce and governance. Any part of a business that currently requires any level of intelligence can benefit from AI. Which
means that every part of the business needs to be AI-enabled.

10
 Taking AI to the next level | Harnessing the full potential and value of AI while managing its unique risks

Taking the next step.

Individual use cases for artificial intelligence are interesting and Whether you are starting with a clean slate and approaching AI from
exciting to think about and implement. However, as AI matures the top down or starting with your existing AI initiatives and building
and becomes increasingly integral to the future of business, up from there, all four dimensions must be addressed. Crafting and
companies need to consider AI’s challenges and opportunities more executing an AI strategy that considers these four dimensions will
comprehensively and strategically. In particular, companies and their help your company scale up quickly to capture the full benefits of AI
leaders need a clear understanding of AI along four key dimensions: without getting bitten by AI’s unique risks.

• Value
Before you know it, you might find your company has become a
• Architecture next gen AI company.

• Workforce

• Governance

Let’s talk
Authors Contributors Contacts
Jeffrey Brashear Erica Macheriotou Amit Desai
Managing Director, Senior Consultant, Principal,
Strategy & Analytics Strategy & Analytics Strategy & Analytics
Deloitte Consulting LLP Deloitte Consulting LLP Deloitte Consulting LLP
Tel: +1 786 907 3615 Tel: +1 212 436 3820 Tel: +1 212 313 2642
Email: [email protected] Email: [email protected] Email: [email protected]

Beena Ammanath Carl Engle

Managing Director, Principal,
Artificial Intelligence Strategy & Analytics
Deloitte Consulting LLP Deloitte Consulting LLP
Tel: +1 415 783 4562 Tel: +1 213 688 4164
Email: [email protected] Email: [email protected]

11
Glossary
Deterministic Symbolic AI Natural language processing
Deterministic AI environments are those on which Symbolic AI is the term for the collection of all Natural language processing, or NLP for short, is a
the outcome can be determined based on a specific methods in artificial intelligence research that are field of study focused on the interactions between
state. In other words, deterministic environments based on high-level “symbolic” (human-readable) human language and computers. NLP helps
ignore uncertainty. representations of problems, logic and search. machines “read” text by simulating the human ability
to understand language. It sits at the intersection
Probabilistic Bayesian inference of computer science, artificial intelligence and
Probabilistic AI environments are those where Bayesian networks are graphical models for computational linguistics.
the outcome is determined based on leveraging representing multivariate probability distributions.
probability-based reasoning and controlling for They aim to model conditional dependence, and Recommendation algorithms
uncertainties to predict outcomes. therefore causation, by representing conditional Recommendation algorithms are those algorithms
dependence by edges in a directed graph. that help machines suggest a choice based on its
Clustering commonality with historical data.
Clustering is a machine learning technique that Connectionist AI
involves the grouping of data points. Clustering Connectionist AI systems are large networks of Knowledge representation
is a method of unsupervised learning and is a extremely simple numerical processors, massively Knowledge representation and reasoning is the field
common technique for statistical data analysis used interconnected and running in parallel. Note: of AI dedicated to representing information about the
in many fields. Clustering is used with applications Connectionism is an approach to AI that developed out world in a form that a computer system can utilize to
including customer segmentation, fast search and of attempts to understand how the human brain works solve complex tasks, such as diagnosing a medical
visualization. at the neural level and how people learn and remember. condition or having a dialog in a natural language.

Categorization Genetic algorithms Sentiment analysis

Categorization is a supervised learning algorithm Genetic algorithms are those algorithms based on Sentiment analysis refers to the use of natural
technique that allows machines to assign categories principles of genetics that are used to efficiently and language processing, text analysis, computational
to data points (categorize data into a given number quickly find solutions to difficult problems. linguistics and biometrics to systematically identify,
of classes). extract, quantify and study affective states and
Virtual agents subjective information.
Anomaly detection A virtual agent is a software program that uses
Anomaly detection (also outlier detection) is the scripted rules and, increasingly, AI applications to Hyperscale cloud providers
identification of rare items, events or observations provide automated service or guidance to humans. Hyperscale cloud providers are those that offer
which raise suspicions by differing significantly from robust, scalable applications and storage portfolio
the majority of the data. Image recognition of services to individuals or businesses to enable
Image recognition is the ability of a system or their architecture to scale appropriately as increased
Regression software to identify objects, people, places and demand is added to the system.
Regression is a statistical approach that estimates actions in images. It uses machine vision technologies
the relationships among variables and predicts with artificial intelligence and trained algorithms to
future outcomes or items in a continuous data set by recognize images through a camera system.
solving for the pattern of past inputs, such as linear
regression in statistics.

Endnotes
1. “Worldwide Semiannual Artificial Intelligence Systems Spending Guide,” IDC, 1H 2019.
2. Jeff Loucks, Tom Davenport, and David Schatsky, State of AI in the enterprise, 2nd edition (survey and Deloitte analysis), Deloitte Insights, 2019.
3. Marco Iansiti and Karim R. Lakhani, Competing in the age of AI,” Harvard Business Review, January/February 2020.
4. b
I id.
5. Gil Press, AI And Automation 2019 Predictions From Forrester, Forbes, November 6, 2018.
6. European Union General Data Protection Regulation

About Deloitte Deloitte refers to one or more of Deloitte Touche Tohmatsu Limited, a UK private company limited by guarantee (“DTTL”), its network of member firms, and their
related entities. DTTL and each of its member firms are legally separate and independent entities. DTTL (also referred to as “Deloitte Global”) does not provide services to clients.
In the United States, Deloitte refers to one or more of the US member firms of DTTL, their related entities that operate using the “Deloitte” name in the United States and their
respective affiliates. Certain services may not be available to attest clients under the rules and regulations of public accounting. Please see www.deloitte.com/about to learn
more about our global network of member firms.

This publication contains general information only and Deloitte is not, by means of this publication, rendering accounting, business, financial, investment, legal, tax, or other
professional advice or services. This publication is not a substitute for such professional advice or services, nor should it be used as a basis for any decision or action that may
affect your business. Before making any decision or taking any action that may affect your business, you should consult a qualified professional advisor. Deloitte shall not be
responsible for any loss sustained by any person who relies on this publication.

Copyright © 2020 Deloitte Development LLC. All rights reserved.