Artificial intelligence (ai) will radically transform the way we do business in the future, and the way we live. That’s a strong statement, but i believe it’s true. Ai has many faces. As we are increasingly exposed to it, it’s important to understand what it can and can’t do and how companies can pivot wisely to this still evolving reality without overlooking the ethical, human and regulatory questions it raises.
3. ARTIFICIAL INTELLIGENCE, A HUMAN REVOLUTION | 3
ARTIFICIAL INTELLIGENCE (AI) WILL
RADICALLY TRANSFORM THE WAY
WE DO BUSINESS IN THE FUTURE,
AND THE WAY WE LIVE. THAT’S A
STRONG STATEMENT, BUT I BELIEVE
IT’S TRUE. AI HAS MANY FACES. AS
WE ARE INCREASINGLY EXPOSED TO
IT, IT’S IMPORTANT TO UNDERSTAND
WHAT IT CAN AND CAN’T DO AND
HOW COMPANIES CAN PIVOT WISELY
TO THIS STILL EVOLVING REALITY
WITHOUT OVERLOOKING THE
ETHICAL, HUMAN AND REGULATORY
QUESTIONS IT RAISES.
4. 4 | ARTIFICIAL INTELLIGENCE, A HUMAN REVOLUTION
WHAT IS ARTIFICIAL
INTELLIGENCE?
In essence, AI technologies sense (capture data),
comprehend (associate a meaning to them), act
(execute an action or pass it on for execution, be that
to a robot or a human) and learn (an essential catalyst
of AI is the ability to learn from historical data in order
to improve its future performance) for a specific activity.
Key in our definition is that we put the focus on the
ability of AI to complement and empower people
instead of replacing them. Therein lies the key: AI is not
omnipotent or capable of replacing us. AI systems are
trained for a narrow, well defined situation. For example,
play chess or Go, understand and handle question or
emails in a certain domain, review and analyze a patient’s
medical information, etc. If you only look at AI from the
perspective of a technology that ‘can do it all’, it will fail
and create tensions within organizations and society.
The added value of Artificial Intelligence lies in its ability
to extend human capabilities. People, not machines, are
at the heart of this so-called Fourth Industrial Revolution.
There is not one but many definitions of AI, and its scope remains fluid
and evolving. Some people even state that AI is everything that has
not yet been done, referring to the observation that as the tools we use
daily become increasingly sophisticated, tasks previously considered
as requiring ‘intelligence’ are now considered routine and get exclud-
ed from the AI definition. Think for example of a good spam filter, spell
check or optical character recognition, all of which used to be consid-
ered revolutionary, but today don’t impress people anymore.
FIGURE 1: WHAT IS AI?
SENSE
Perceive the world
by acquiring and
processing images,
sounds and speech.
COMPREHEND
Analyze and under-
stand the informa-
tion collected by
adding meaning and
insights.
ACT
Take action in the
physical world
based on compre-
hension and under-
standing.
LEARN
Improve performance
(quality, consistency,
and accuracy) based
on real world expe-
riences.
Accenture defines AI as follows:
“A constellation of technologies that extend human capa-
bilities by sensing, comprehending, acting and learning –
allowing people to do much more.”
5. ARTIFICIAL INTELLIGENCE, A HUMAN REVOLUTION | 5
1. Computing power continues to grow and
decrease in cost, making it possible to capture
vast volumes of data and run increasingly
complex machine learning.
2. Data is growing exponentially and the Internet
of Things and Big Data solutions are creating
new data sets each day. The resulting ocean
of data provides the ideal basis for training AI
tools.
3. AI software packages and toolkits are
becoming increasingly available, often
offered as plug & play solutions through the
Cloud.
4. Lastly, the open innovation aspect of AI
means it is easier to gain access to leading
AI thinking and skills. As a result, more and
more people will be able to create a bridge
between humans and these technologies.
WHY IS AI CATCHING ON
NOW?
While AI exists since the 1950s and a good number of the AI tools we
use today have been around for a while, its adoption has grown ex-
ponentially in the last couple of years. There are multiple reasons for
why AI is affordable, doable and available today:
6. 6 | ARTIFICIAL INTELLIGENCE, A HUMAN REVOLUTION
1956
When AI first emerges as an ac-
ademic discipline at Dartmouth
University in the US, people mar-
veled at the ability of computers
to play checkers, solve algebra
problems and speak English.
1960s
AI research is heavily funded by
the US Defense Department and
its founding fathers are optimis-
tic, predicting “machines will be
capable, within twenty years, of
doing any work a man can do”.
1974
Research funding stops, leading
to a period that becomes known
as AI Winter.
Early 1980s
AI research is revived due to the
success of expert systems that
emulate the decision-making
ability of a human expert.
Late 1990s
Increased computational power
sees AI used in, amongst other
areas, data mining and medical
diagnosis.
1996
IBM’s Deep Blue defeats World
Chess Grand Master Garry
Kasparov.
2011
IBM’s Watson defeats two former
human champions at the quiz
show Jeopardy!
2016
Google’s DeepMind AlphaGo
defeats legendary player Lee
Sedol in the ancient and highly
complex Chinese game of Go.
January 2017
The AI computer program Libra-
tus challenges 4 professional
poker players to a 20-day compe-
tition playing 120,000 hands, and
ends over $1,7 million in chips,
ahead of its human rivals.
October 2017
Google’s DeepMind AlphaGo
Zero learns to play Go on its own
without human help.
AI is not a new technology, but it is
now evolving fast:
7. ARTIFICIAL INTELLIGENCE, A HUMAN REVOLUTION | 7
AI is not an automation tool that allows to improve
productivity and reduce costs. Companies need to truly
consider it as a ‘new factor of production’ which offers
them opportunities to augment their human labor and
physical capital and by doing so drive their growth,
profitability and sustainability.
The use of AI can allow companies to create richer,
hyper-personalized experiences, hence driving
customer satisfaction and retention. It can help to better
understand and predict customer needs; and based
on that develop innovative business models, products
and services. All these leverage have the potential to
generate additional top line growth.
At the same time, the obvious impact on profitability
through increase productivity, error reduction and
improved quality and flexibility in capacity is not to be
ignored.
Less obvious, but not less valuable is the ability of AI
to help build sustainability and trust with client, own
employees and regulators.
Starting from those three levers – growth, profitability,
trust – the business value of AI can be structured around
intelligent automation, enhanced judgement, enhanced
customer interactions and enhanced trust – each time
enabling people and businesses to grow and do more.
HOW TO USE AI?
Used in the right way, AI can help drive growth, profitability and
sustainability.
Illustrative increase to enterprise
value in 10 years.
FIGURE 2: HOW TO USE AI?
APPLYING A BALANCE OF AUTOMATION AND AUGMENTATION
Value Today
INTELLIGENT
AUTOMATION
10-20%
15-25%
10-20%
20-30%
10-15%
Value with AI
ENHANCED
JUDGMENT
ENHANCED
INTERACTION
INTELLIGENT
PRODUCTS
ENHANCED
TRUST
8. 8 | ARTIFICIAL INTELLIGENCE, A HUMAN REVOLUTION
INTELLIGENT AUTOMATION
Adding cognitive capabilities on top of automation
technologies enables them to be self-learning and to
act autonomously and proactively. The benefits not
only include cost reduction, more efficient processes,
activities and services but also error reduction,
increased quality, consistency and flexibility in capacity.
ENHANCED JUDGMENT
AI solutions are able to ingest and digest larger quantities
of data than any human is capable of. As such, they
can be used to augment human intelligence in human-
driven processes such as strategic decision-making. By
collecting, organizing, analyzing and presenting data
for human judgment they enable people to do a better
job, while supporting decision ‘explainability’.
ENHANCED INTERACTION
AI powered virtual assistants help businesses to deliver
a superior customer experience based on natural
language dialogues, hyper-personalization and the
curation of real-time information. This can increase
customer acquisition, retention and overall satisfaction.
INTELLIGENT PRODUCTS
AI is being applied to introduce a new class of innovative,
intelligent products and services, at high speed and of
high quality. These products capture data from their
real-life environment and use that data to steer the
characteristics and behavior of the product itself and/
or its pricing structure. Examples include Self-Driving
Cars, Lighting as a Service, Industrial Internet of Things
Self-Repairing Software, and Contextualized Insurance.
ENHANCED TRUST
Trust has become a key challenge in the digital age.
AI solutions can be used to improve governance,
compliance and transparency both within an
organization and outside, with external stakeholders.
9. ARTIFICIAL INTELLIGENCE, A HUMAN REVOLUTION | 9
With AI covering a broad range of very different techniques, the question around maturity does not always have
a simple answer. The fast paced evolution in technology solutions available on the market further increases the
challenge of getting a clear view on what is ready for use already today; and where some caution is advisable
Based on our hands-on involvement in AI pilots and roll out projects at different clients, we have created an
overview of AI techniques which have proven both their value and their implementation feasibility in a real life
context. Below an overview of the four families of AI techniques which are the most actively explored:
In real life, these techniques are often combined to get to optimal results: email assistants and chatbots are trained
through machine learning to raise their performance; computer vision techniques are added to an email assistant
to be able to capture additional data from stamps, signatures and handwritten comments on documents, and so on.
Bringing AI to action will mean finding the right use cases and the right techniques – in an ever-evolving landscape.
AI IN ACTION
Select and combine techniques to improve human performance.
Natural Language Processing
An AI system that understands and uses language as used by humans without imposed
structures, words or commands. For example, an intelligent email assistant can
process incoming emails 24/7, in multiple languages, increasing internal efficiency
and improving customer experience. (see page 8).
Chatbots & virtual assistants
Natural language dialogue based and machine learning enabled, these computer-
generated ‘characters’ can converse with people and answer their questions,
transforming the way of interacting with customers. For example, Accenture and Fjord
have developed a virtual mortgage agent, Collette, to provide personalized advice on
mortgage products to individuals in the UK (see page 10).
Machine learning
By leveraging data and experience to improve its performance, supervised,
unsupervised and observational machine learning is used to handle large and
complex data sets, for example in fraud detection, claims underwriting, credit scoring
and micro customer segmentation. In the latter, the ability to analyze internal and
public data on customers through powerful algorithms drives customer acquisition
and profits.
Computer vision
AI Systems that are able to understand and extract meaning from digital images
and videos are often used in safety and security monitoring, for example, assessing
suspicious behavior, monitoring traffic, detecting abandoned objects, etc. this
AI technique is however also finding its way into other industries, for example in
assessing car damage for insurance claims, which is currently a lengthy process
involving multiple intermediaries.
10. Natural Language Processing (NLP) uses natural
human language in interactions between computers
and people, adhering to the new paradigm of systems
adapting to humans instead of the other way around.
An NLP system can undertake one of two tasks:
• Natural Language Comprehension: Extracting
meaning from spoken/written language;
• Natural Language Generation: Generating a
message in a natural (human) language, for example
contracts and meeting minutes.
The job of the NLP powered intelligent email
assistant is to process incoming emails and their
attachments, for example, emails asking questions
about a bankcard, a loan or insurance claim, and
provide information and/or take follow-up actions.
All four of the defining features of AI are present: the
system senses the content in the email by receiving the
email and attached documents, and transforming them
into workable text. It then comprehends the content for
example by categorizing the email based on the topic,
by extracting specific data and/or performing sentiment
analysis. It will then act, based on its comprehension of
the email content, to define the required action, and to
either execute that action itself or forward the email to a
human or another robot for further follow-up. Finally, the
intelligent email assistant will continuously learn from
past events to improve the accuracy and correctness of
the email handling solution in the future.
Like humans, such systems can’t deliver 100%
accuracy. Their performance depends on the
linguistic complexity of the emails, the breadth of
the categories and data fields they need to extract,
as well as the quality of the written texts they are
dealing with, e.g. very short extracts can be harder to
understand and language can often be ambiguous.
The availability of a sufficiently large and clean historical
data set often proves to be a challenge, slowing down
the initial machine learning-based training of the NLP
tool. However, with time and continuous training an
NLP powered intelligent email assistant can achieve
up to 80%-90% ‘recall’ or completeness (the number
of emails handled automatically by the system) and
70%-90% ‘precision’ or correctness (the number of
emails handled automatically with the right outcome
– for example classified in the correct category, or all
required data correctly extracted).
The NLP powered intelligent email assistant
CASE 1
SENSE
Receive email and
transform data into
workable text
COMPREHEND
Understand email content
(and attachments) for example
categorize email, extract data,
define sentiment
LEARN
Leverage machine learning
techniques to learn and improve
the accuracy and correctness of the
email handling solution
ACT
Based on the comprehension of the email
content, define required follow up action
- Define priority
- Create entry in ticketing tool
- Assign to correct agent or Robot for
follow up
SKILL A
SKILL B
RPA
Manual
Follow-up
Manual
Follow-up
Automated
Follow-up
CUSTOMER INTELLIGENT EMAIL ASSISTANT SERVICE DESK AGENT
THE NLP POWERED INTELLIGENT EMAIL ASSISTANT
FIGURE 3: THE NLP POWERED INTELLIGENT EMAIL ASSISTANT
11. ARTIFICIAL INTELLIGENCE, A HUMAN REVOLUTION | 11
Accenture uses an
Intelligent Email Advisor
to manage its own biggest
inbox
With more than 435,000 employees worldwide
requiring support on 2000+ applications and
500,000+ devices, accounting for more than
1,5 million interactions per year, the Accenture
Technical Helpdesk sees its fair share of emails
coming in. By augmenting the human agents
through an AI powered Intelligent Email Advisor
system able to understand 22 languages, low
complexity tasks have been automated, freeing
human operator bandwidth to focus on value
added tasks and providing an overall faster and
more convenient user support.
The Intelligent Email Advisor captures data
from the email sender, subject, email body
and attached documents and uses that to
understand and log requests. It proceeds to
handle and close simple requests itself, and
passes more complex requests and emails
in which it detects dissatisfaction through a
negative tone of voice to a (human) IT associate
for manual handling.
The performance of the email handling set-up
is continuously improved based on feedback
provided by the IT associates and suggestions
made by the system itself – suggestions that
require validation by the IT associate before
they can be applied.
12. 12 | ARTIFICIAL INTELLIGENCE, A HUMAN REVOLUTION
Collette is an AI powered mortgage adviser that
combines the ability of interactive dialogue, natural
language comprehension, supervised machine learning
and advanced analytics to provide customers with
tailored advice on the mortgage best suited to their
needs and circumstances. Drawing upon an extensive
corpus of content knowledge about mortgages, the
chatbot can understand the intentions and needs of
the customer, inform him/her on the implications of a
certain choice and offer access to additional help on
demand (for example explanatory videos) when needed.
Collette not only significantly reduces the time-
consuming pre-mortgage assessment process and
frees up human mortgage advisors to focus on
customers with more complex needs. It also provides
customers with an additional means for interaction
with their bank which is available 24/7 and which
for some customer segments is better aligned with
their preference for digitally enabled communication.
How Collette works
Collette plays its advisory role as part of a broader end-to-
end mortgage purchase process; offering customers at
any point in the process the choice to use Collette, or not.
1. A customer applies for a mortgage online and
can opt to start a chat with Collette - the virtual
mortgage adviser – to get further advice.
2. Collette uses guided and unguided dialogues
to conduct a meaningful conversation with
the customer covering important topics
such as their financial situation, the type of
mortgage they want, rates, fees and payment
schedule.
3. As part of the conversation, Collette will ask
customers to explain their decisions, this
to validate that they have fully understood
the concept and the consequences of the
choices they are making.
4. Multiple levels of ‘help on demand’ are
available, for example in the form of
embedded videos, for customers requiring
more detailed information.
5. The customer can at any point decide to
abandon the chat, or Collette herself can
opt to refer the case to a human agent
when required, for example when she does
not understand the customer input, or for
customers with more complex needs.
6. Based on the collected data, Collette presents
the customer with a tailored mortgage
recommendation that fits their needs and
circumstances.
Collette: The Mortgage Adviser of the Future
CASE 2
Watch Collette in action:
youtube.com/watch?v=uXUh-aFO_bA
IMMEDIATE 24/7
SERVICE, ENHANCING
CONVENIENCE FOR
CUSTOMERS
ALLOW MORTGAGE
ADVISORS TO FOCUS
ON CUSTOMERS WITH
COMPLEX NEEDS
COMPLIANT WITH
MORTGAGE CONDUCT OF
BUSINESS (MCOB)
RULE BOOK
COST PER ADVICE
SESSION REDUCED BY
C. 95%
FIGURE 4: COLLETTE’S BENEFITS
13. ARTIFICIAL INTELLIGENCE, A HUMAN REVOLUTION | 13
UK regulatory change as a
trigger for the creation of
Collette
Following the 2007 – 2010 subprime mortgage
crisis, the UK Financial Conduct Authority
(FCA) introduced in April 2014 a new set of
rules called the Mortgage Market Review
(MMR). To comply with this new regulation,
most customers must now receive formal
advice from their bank before taking out
a mortgage. The exponential increase in
workload this creates for banks not only leads
to a challenge in scaling up and financing their
mortgage advisory teams, it also significantly
lengthens the mortgage purchase process
for customers, burdening the quality of their
interactions with their bank.
The creation of Collette as a virtual mortgage
adviser offers a means to scale a lot faster and
at a lower cost, while at the same time offering
customers an innovative, digital way of
interacting with their bank at any time it suits
them. Collette therefore solves the immediate
industry issue resulting from the introduction
of MMR while at the same time transforming
the credit services industry, independent of
the original regulatory trigger.
14. 14 | ARTIFICIAL INTELLIGENCE, A HUMAN REVOLUTION
There is no denying the added value AI systems can
bring. At the same time, we must not ignore their power
and with that, our responsibility to use them ethically.
This means making sure people remain at the center
(enhancement of human activities, new skills, re-
training…); that algorithms are not discriminatory (for
example, in loan or insurance decisions); that personal
data is protected; and labor and employment laws are
complied with.
The secret is to think wisely, experiment and collaborate.
But beware: this will trigger a shift at the core of the
organization that will impact the way of doing things
forever.
With great power comes great responsibility:
the need for Responsible AI
HOW TO BEGIN?
AI PROMISES GREAT OPPORTUNITY, AND WITH THAT COMES GREAT
RESPONSIBILITY FOR GOVERNMENT AND ENTERPRISE LEADERS ALIKE
Design solutions to benefit clients and
employees
Integrate human intelligence with
machine intelligence
Plan for reskilling of employees who are
impacted by AI
Ensure fairness and adherence to
company core values and ethical principles
Build for transparency and explainability
Eliminate algorithm and data bias
Ensure clarity on human versus algorithm
accountability
HUMAN AT
THE CENTER
ETHICAL
DESIGN
REGULATORY/
POLITICAL COMPLIANCE
Adhere to regulation – eg data privay
protection, IP ownership
Evolve AI deployments in line with evolv-
ing domestic and regional regulation
Influence and advise regulators based on
experience
FIGURE 5: RESPONSIBLE AI
15. ARTIFICIAL INTELLIGENCE, A HUMAN REVOLUTION | 15
Today, leading businesses are
using AI to transform their core
business—creating completely
new revenue streams. So what are
AI leaders doing that’s different?
And how can you learn from them
to boost your AIQ?
To turn AI investment into AI-driven growth, companies
will need in-house, proprietary capabilities for AI:
they will need to own some of the talent, some of the
technology, and some of the data. They will also need
to be deeply involved in a broader ecosystem. Neither
startups nor incumbents will thrive with a ‘not invented
here’ approach. Open innovation combines the muscle
and maturity of incumbents and the ideation and agility
of startups.
A recent Accenture study1
of the Fortune Global 100
and what we call the Intelligent Global 100 – pioneers
in the development of AI applications and technologies
– introduces the ‘Artificial Intelligence Quotient’ (AIQ)
concept to measure the AIQ of a company by looking at
their in-house focus (AIQ for invention) and their outside
focus (AIQ for collaboration).
Less than 20% scored well on both indexes. These
‘collaborative inventors’ have realized significant
growth in enterprise value (4.3% since 2013
compared to 2.3% on average for the rest of our
sample), which shows that AI is already generating
returns today.
56% - companies we call ‘observers’ - were weak on
both indexes and still adopt a wait & see approach,
in our opinion leaving a significant enterprise value
increase opportunity untapped.
1. ConductedbetweenJanuaryandMay2017,AccentureundertookthisresearchandanalysisonbehalfoftheG20YoungEntrepre-
neurs’Alliance.https://www.accenture.com/t20170614T050454__w__/us-en/_acnmedia/Accenture/next-gen-5/event-g20-yea-summit/
pdfs/Accenture-Boost-Your-AIQ.pdfAccentureMarketPulseSurveyBelgium2017.
DO YOU HAVE A HIGH
AIQ?
Learn from the AI leaders
16. 16 | ARTIFICIAL INTELLIGENCE, A HUMAN REVOLUTION
COLLABORATIVE
INVENTOR
17% OF COMPANIES
1. Use AI to transform core business
2. Develop AI in-house, benefit from
owning AI-critical resources
3. Collaborate to share resources and to
co-create AI
INVENTOR
14% OF COMPANIES
1. Use AI to transform core business
2. Develop AI in-house, benefit from
owning AI-critical resources
3. Only collaborate to source talent
COLLABORATOR
13% OF COMPANIES
1. Use AI to drive incremental value
across business
2. Collaborate to adopt AI solutions and
services, relatively small in-house
inventions
3. Unable to own AI-critical resources or
limited benefits from their ownership
OBSERVER
56% OF COMPANIES
1. Do not fully see transformational or
incremental value of AI
2. See limited benefits in owning AI-
critical resources
3. Relatively small initiatives, wait and
see approach
AIQ FOR COLLABORATION
AIQFORINVENTION
KEY Collaborative Inventors Other companies
FIGURES 6 & 7: AIQ MATRIX FOR INVENTION AND COLLABORATION
17. ARTIFICIAL INTELLIGENCE, A HUMAN REVOLUTION | 17
AI MUST BE HANDLED
WITH CARE...
AS WE REACH THE NEXT LEVEL OF AI
TECHNOLOGIES, WE MUST ALSO BRING
BUSINESSES AND PEOPLE TO THE NEXT LEVEL,
EMPOWERING THEM AND BUILDING A BETTER
FUTURE FOR SOCIETY. THERE ARE NO IDEAL
RULES OR TIMESCALES FOR DOING THIS. PIVOT
TO THE NEW WISELY.