Professional Documents
Culture Documents
Free Ebook The Role of AI in Finance 1712162301
Free Ebook The Role of AI in Finance 1712162301
of Finance:
01 Acknowledgments 2
Executive Summary 3
introduction 4
02 AI - back to basics 6
11 Sources 41
01
Acknowledgements
Part 01 Acknowledgements
Executive Summary
Introduction
CFOs and other finance leaders
have plenty on their plates in 2023.
Gone are the days when all they had to worry about
were balance sheets, P&Ls and management reports.
Today’s finance leaders are operating in volatile and
uncertain times. Ankur Agrawal from McKinsey
recently wrote about the blend of opportunity and
challenge that modern CFOs face.
than ever - but with their wider roles, there is also less time for finance leaders and
CFOs to do so. This is part of the reason the industry is so excited about AI.
AI tools can analyze vast quantities of data, and have the power to identify trends
that the human eye could easily miss. That means better insights, more accurate
reports and forecasts and better-informed strategic decisions. Also, AI tools can
cut a swathe through routine tasks like report creation and even customer
enquiries, leaving human brains free to focus on more interesting and rewarding
value-adding activities.
AI - back to basics
Before we dive too deep down the rabbit hole,
it is as well to be clear what we are talking about
when we discuss Artificial Intelligence.
We think of AI as a disruptive technology, and rightly so, but that does not mean it
is conceptually new. It is simply the speed of development that is liable to catch
businesses napping.
So what do we mean by AI and how can it help us reach our desired destination or
achieve our strategic goal? In short, any software or program that exhibits what we
would call intellectual quality that we would associate with cognition is an example
of AI. So in other words, this is software that can learn, solve problems, understand
From a business perspective, we can immediately draw out some specific examples
of AI in action:
the basis of what it “knows” from existing data and past experience.
language and interacts with the data. Software that automatically reads and
In this era of Siri and Alexa, we encounter AI every day. However, the somewhat
scripted responses seem more “A” than “I”, using simple voice-recognition NLP to
select an appropriate resource. We actually encounter far more powerful AI, but in
When you browse your social media news feed or Google Discovery over breakfast,
the news stories at the top are all on the topics that interest you. It’s not by
coincidence, AI uses your browsing history to anticipate your needs and interests.
Similarly, navigation apps learn our daily routines so they can recommend the best
time to leave for work and what route to take.
The latter is a particularly pertinent example as it shows how AI can analyze data
and make the best recommendation, saving humans from having to work it out for
themselves. In this case, it is quite a simple process based primarily on the latest
traffic reports on different roads. But this kind of decision-making extends far
beyond the preserve of handy consumer gadgets.
AI can offer both faster and more reliable decision-making. Faster because it can
analyze thousands of data records in a second and more reliable because it can
therefore base its recommendations on a far larger data set. The implications are
significant. Take medicine as an example.
A doctor might spend an hour consulting a dozen similar cases when attempting to
diagnose an unusual case. AI can analyze thousands of cases much faster. Not only
does it mean a quicker and more accurate diagnosis, it frees up an hour of the
doctor’s time that can be spent treating patients.
Key takeaways
How generative
AI helps finance teams
There are more similarities than might be immediately obvious between the doctor
in our previous example and the average CFO or senior finance manager.
We mentioned earlier that finance leaders have a broader remit than ever before.
The right AI tools can take routine tasks from their shoulders, effectively
transforming their day to day activities into more supervisory and strategic roles.
There might be a limited application for this kind of “digital CFO” for smaller
businesses and tech-savvy startups, but as Kevin Held, the CFO at Hazeltree, said in
early 2023, finance and accounting teams have evolved into key business partners,
overseeing areas such as transparency in cash flow and other strategic needs.
AI can be used to increase the speed and confidence of decisions in three ways:
Decision automation
The generative AI system makes the decision on the basis of either
prescriptive analytics or predictive analytics. This sort of decision-making is
being used in insurance technology (insurance tech) to dramatically speed
up the payout process, for example.
Decision augmentation
This works on the same basis as automation, but the system stops short of
acting on its decision. Instead, it presents a recommended course of action,
or a shortlist of options, to the human user, who can then make the final
decision, backed up by the significant analysis that has been performed by
the AI.
Decision support
The AI provides data-driven insights and the user considers these alongside
human knowledge, experience and “gut feel” to arrive at a solution.
The businesses that can make the best use of the tools at their disposal and
understand which type of automation, augmentation or support is most appropriate
are the ones that will leverage a true competitive advantage.
Finally, when looking at how generative AI helps finance teams, let’s not overlook the
simplest but often most impactful aspect. We talked dismissively of “number
crunching” earlier because it is something software can do so easily. But automating
any task can be hugely beneficial in terms of efficiency, accuracy, and even
job satisfaction.
A mechanical digger can shift as much earth in an hour as would take a human
worker 40 days, and it would make a tidier job of it. Data preparation is surely the
finance equivalent of shoveling earth, and offloading it onto an automated system is
an innovation against which few would push back. It is just one example of workflow
automation that can deliver major benefits. Companies from Coca Cola to Netflix
have used AI for similar simple but profoundly effective automation tasks such as
analyzing customer feedback or providing new product recommendations.
Key takeaways
One of the most profound aspects of AI is that once it starts down a particular
the chess computers that caused such a stir in the 1980s and 90s.
competition and won. In 1997, it beat Garry Kasparov, the greatest chess player
the world had ever seen. In fact, Deep Blue and Kasparov were quite evenly
matched, but today’s chess computers like Stockfish are vastly superior to both
Deep Blue and any human chess player. In fact, the last time a human beat a
computer at chess was 15 years ago, and it won’t ever happen again. Stockfish
might think 10 to 12 moves ahead. It’s a little like comparing our guy with a shovel
Of course, a chess computer is a very narrow example. There are more possible
variations in a chess game than there are atoms in the observable universe, but
it is still a finite and bound set that conforms to specific rules. Code those rules
and the AI can just go deeper and deeper down the rabbit hole. Richard Branson
is famous for his sometimes fanciful comparisons between the two, but the truth
In particular, the sheer volume of data that is now at every business’s disposal is
every bit as mind-blowing as the 10111 moves that are possible on a chess board. It
is at this point that you start to appreciate that any apparent similarities between
If you feed garbage in, you will get garbage out, and you would do as well to ask
your cat or your lucky 8-ball to make the decision.
We will explore that angle further in just a moment, but the point to understand
right now is there is a big gulf between raw data and insights. It is interesting to
note that technopedia defines big data not as a noun, but as a process by which
“a massive quantity of data can be converted into useful information.” Not only is
this data vast in quantity, it is also of a whole new type compared to that which
analysts have used in years gone by.
Businesses are accustomed to dealing with structured data, of a kind that can
easily be tabulated in a spreadsheet or a text file. Much of the data that is
generated today, however, is unstructured. For example, emails, social media
shout outs or comments, Google reviews, audio or video files, they all contain
data that can be immensely valuable.
Then there are diverse data sources to consider. The Internet of Things has
dramatically expanded the potential here. Smart phones, wearables, even data
from vehicles and smart road networks can all come into play, depending on the
type of business.
All of these demand effective big data processing capabilities if they are to be of
any value in delivering useful and actionable insights. Otherwise, there is a risk of
expending vast efforts and resources into what turns out to be nothing more than a
resource drain.
We could talk for hours on the topic of data processing and analytics. For our
present purposes, the point is that it is one thing to conceptualize how the data
needs to be processed and analyzed. But to put the concept into action demands
the power of AI. Again, we are back to the earth moving equipment carrying out a
task that someone with a shovel can only imagine.
Unlike the human brain, AI is not affected by personal bias. It does not forget things
or get distracted or have an off day. Before AI, there was an upper limit to the
quantity of data that could efficiently be processed by the human brain. Going
deeper or bigger became so resource-intensive that the return was simply not
worth the investment. With an AI-based system, the sky is the limit. In fact, as we
have already discovered, the more data it is exposed to, the more insightful
the results.
Best of all, once AI is up and running and aligned to our objectives, it can handle its
own fine tuning. In the past, performing traditional manual data analysis
demanded a whole new set of code every time there was a change to the
parameters or requirements. AI can perform whatever adjustments are needed
independently, essentially working on a constant basis of self-improvement.
Key takeaways
You can have all of the fancy tools, but if your data
quality is not good, you're nowhere.”
It sounds tongue in cheek, but the concept is deadly serious. When investing in AI to
analyze data, you need to be absolutely certain you are giving it the best possible
data to analyze. It sounds obvious, but as we have already discovered, life and
business are more complex than a game of chess.
One thing businesses are great at in the 21st century is coming up with fantastic
solutions and then desperately looking for problems they might be able to solve.
That’s a trap that is so easy to step into in data analytics. The mere existence of
data is often deemed sufficient reason to feed it into big data then sit back and wait
for the lightning insights to rain down.
Sad to say, it doesn’t work that way. Grant Thornton developed a useful
framework that helps you identify what data is useful by allocating it to one of
Stakeholder management
Improving stakeholder
Value creation
Delivering strategic
initiatives to
create additional
Operational
shareholder value
delivery
Improving efficiency
and accuracy in
operational delivery
Business protection
Protecting your
If data provides insights in one of these four areas, then it is data that is worth
looking at. The model also provides an opportunity to identify data that could be
of future value and can be placed on the back burner if it serves an area that is
Once you are clear about what you are collecting, the next area to address is how
you are collecting it. Today, most businesses use data pools for this purpose. Put
simply, a data pool is a centralized repository where data is housed from diverse
sources. Don’t confuse a data pool with a data warehouse. The latter stores
historical data that can be interrogated for statistical and analytical purposes. The
former is a living and breathing thing that houses real time data, or at least as
close to real time as is practically possible.
Data pools are invaluable in dynamic business environments where decisions need
to be made on the hoof and based on the latest market information. As well as
providing the basis for which AI can operate, they also serve to foster and enhance
cross-team collaboration within an organization by providing a single and
accepted source of truth. For years, teams and business units have operated within
their own siloes. Having this sort of information shared and accessible to all helps
humans to make better decisions and operate more efficiently as well as AI.
Key takeaways
Pigment AI
Part 06 Case study Pigment AI
Pigment AI
The foregoing all makes plenty of sense in theory,
from the single data pools shared across the
business to the AI software with an insatiable hunger
to learn. But how does it all look in practice in a real
world business? Pigment AI was recently announced,
giving finance teams a glimpse of the future.
The people at Pigment set out to deliver a solution that adds genuine value while
avoiding the need to reinvent any wheels. The enterprise-grade platform
integrates with a business’s existing tech stack to provide a single data pool for all
the company’s data. The AI component opens up a world of new reports,
information, trends and insights. It’s fast, it’s comprehensive, and it includes a user
interface that makes it simple to use. It is also highly flexible to meet the diverse
needs of different clients.
Pigment AI does not set out to obviate the need for decision making by placing it in
the hands of computer software. On the contrary, it describes AI as a “trusted co-
pilot” that can provide the insights that enhance both decision-making and
productivity. It does this by providing instant answers to the right questions that
allow CFOs, finance heads and other decision-makers to focus on those activities
that have genuine strategic benefit and ultimately add value to the business.
The specific deliverables of Pigment AI will, of course, vary from one use case to the
next. However, they can be summarized in the following terms:
1
Employees across three business can onboard and get working with
their own models to get value out of the software without delay
2
Pigment has all the information that’s needed for AI to deliver the
in a format that is easy to grasp and digest. It will even write a concise
3
Pigment doesn’t quite come with its own crystal ball, but it can use the
4
Delivering actionable insights is one thing. Delivering action is another.
and action points and sharing these across the workforce so that
strategic goals.
Examples and
some food for thought for any CFO asking “what can
AI do for me?”
What’s more, the phrase “losing the will to live” was used advisedly. Every CFO was
once a part-qualified trainee and will know that those repetitive tasks added
nothing to the employee experience. Today’s starters can avoid that particular rite
of passage and instead get involved in more rewarding activities sooner.
Even the greatest humans make mistakes all the time, so we can’t be surprised if a
data admin clerk on a minimum wage working overtime slips up occasionally. In
2019, the State of Pennsylvania overpaid a school district by half a million dollars
due to a typo on a spreadsheet. Not only can AI be trusted not to blunder, it is also
highly adept at spotting any outliers from human errors at the granular level - an
example of garbage in not necessarily leading to garbage out.
The changing nature of the finance role that AI is making possible offers up new
scenarios. Suddenly, employees from traditionally underrepresented backgrounds
or completely different areas of the business have the chance to to enter or
advance within the finance function without having to dedicate five or six years to
night school or an elearning degree - unless they want to, of course. It provides
enriching opportunities that work both ways.
Paradoxically, while both AI and the finance sector face a degree of suspicion in the
public eye, their coming together could be exactly the catalyst that is needed to
restore trust. It all comes down to AI being used for the public good, and just as
importantly for it to be seen to do so.
We have seen in the preceding pages that leveraging the power of AI can certainly
change financial services for the better, benefiting businesses, their stakeholders
and finance managers themselves. But those benefits also extend to customers
and consumers in B2C businesses. AI can make businesses more responsive and
can tackle fraud and financial crime, which ultimately benefits everyone except
the criminals.
AI certainly has the power to restore trust in financial services where it is currently
lacking. But only if it can be trusted and accepted in itself. Where that is concerned,
it is partly down to businesses and specifically finance teams to use AI tools to add
value for customers and bring opportunities to personnel.
Transparency is always vital for gaining trust. People have a natural suspicion of
change, and that applies to middle-management accountants as much as it does
to consumers who entrust their personal and financial data to businesses. In both
cases, it is vital to clearly articulate what the new tech is, what it isn’t and what
benefits it delivers. If its value is not clear, then it begs the question of why it is being
introduced in the first place, so this process provides a valuable sense-check, too.
Key takeaways
Yes, AI is a tool, but it is a fundamentally important one that will touch everyone. A
little like the broader IT revolution of the 1990s, this is a journey that everyone needs
to take, not just those in IT or finance. Businesses tend to focus reskilling on the
highest performing employees, when really they need to think first about those who
are most at risk - one last reminder of our friends with the shovels.
Research by the WEF shows that those with the greatest need for reskilling are
commonly the least likely to get it. Don’t let your business fall into this trap.
Key takeaways
2 Prioritise those who are most at risk, not just the top performers
complete AI glossary
as possible.
Data science: Any field of research that involves processing large amounts
types
Machine learning (ML): Allows computers to learn from a set of data points
Th
T he
e F
Fuut
tuur
ree o
off F
Fiin
naan
ncce
e:: E
Exxp
pllo
orriin
ngg t
thhe
e R
Rool
lee o
off A
AII 40
4
11
Sources
Part 11 SOURCES
Sources
mckinsey.com
In conversation: The new CFO mandate
sage.com
AI in finance: What ChatGPT and generative AI means for CFOs
gartner.com
How to Make Better Business Decisions
freepctech.com
What Now for Humans As Computers Get the Upper Hand At Sport
flowster.com
AI Workflow Automation: How it Can Improve Your Business Processes
liverpoolmuseums.org.uk
Which is greater? The number of atoms in the universe or the number of
chess moves?
influencermarketinghub.com
Using AI in Data Analysis – How to Make Sense of Big Data
grantthorn.co.uk.com
Data quality: how to avoid 'garbage in-garbage out'
eu.poconorecord.com
State overpaid Sburg $500K
fca.org.uk
The future of regulation: AI for consumer good
techtarget.com
What is generative AI? Everything you need to know