Ai 1

Changing the game:
the impact of artificial

intelligence on the banking
and capital markets sector
Contents
Overview: where are banks in the AI journey? 01
What impact can AI have on the bottom line and how? 04
How will the sector landscape change and who will be successful? 10
What is special about generative AI and where is this technology heading? 14
How will banks embed AI across the value chain? 17
What risks must be managed and how? 23
What are the key considerations for safe and effective execution? 26
How to get started, scale and drive adoption 28
Contacts 31
Endnotes 33
Changing the game: the impact of artificial intelligence on the banking and capital markets sector
1. Where are banks

in the AI journey?
01
1. Where are banks in the AI journey?
Artificial intelligence Artificial Intelligence (AI) is already here and

shaping the wider world banks operate in.
Considering the sector outlook more
generally, the coming years continue to
will likely determine the In automotive, Tesla and others delivered include macroeconomic and geopolitical
AI technology for sophisticated driver-assist uncertainty. Any number of unforeseen
banking and capital functions, with an eventual end goal of events may emerge from an already cloudy
markets sector’s autonomous vehicles operating on public

roads.1 The life sciences industry has been
crystal ball. However, in a five-year timeline,
we in Deloitte Global Financial Services
winners and losers realizing value from AI for drug research see AI as the single biggest controllable
and new molecule discovery, as it can draw opportunity for players to improve their
in the coming five insights from massive data sets faster, competitiveness.
years. The journey process data and automate workflows

more efficiently, and convert insights into AI now allows banks to tackle challenges of
has already started. actions to improve business performance scale in a way that, previously, would have
– from molecule to market.2 In public required many extra staff. If a particular
safety and security, for example in the function in a bank could be done better or
United Kingdom, London’s Metropolitan faster by adding one hundred extra trained
Police has trialled live facial recognition staff, it’s likely that AI can be transformative
(LFR)3 cameras in specific areas, to for that function. AI offers vast additional
accelerate identification of individuals the operational capacity, at low marginal
police are looking for. Regulating for the cost compared to hiring the equivalent
evolving use of AI is an ongoing challenge processing capacity as staff.
to lawmakers, for example the European
Union’s AI Act is intended to protect health, But more than that, the game in which
safety, fundamental rights, democracy players are competing will likely change.
and the rule of law, and the environment AI is on the threshold of a paradigm
from potential harmful effects – while shift. Through the work we do with
supporting innovation, particularly banks around the world we see leading
among European SMEs (small and innovators already making the step from
medium enterprises).4 AI as an ‘instrument of strategy’ (i.e.,
accelerating delivery of today’s business
Within this evolving societal context, AI is plan) to a ‘determinant of strategy’, where
not new to the banking and capital markets tomorrow’s business is planned around
(B&CM) sector. It has been in production new AI capabilities. JP Morgan Chase, which
for years in specific functions, including topped Evident Insights AI Index (which
algorithmic trading and trade surveillance. benchmarks how ready banks are for the
But the arrival of Generative AI (GenAI) incoming wave of transformation that
marks a new era, exploding the number of AI will bring) for a second year5, sees the
potential use cases and putting benefits in transformational impact that AI can have
the hands of the workforce. and plans to spend $1 billion or more
a year on AI capabilities.6
AI now allows banks to tackle

challenges of scale in a way that,
previously, would have required
many extra staff.
02
An important point is that we do not see

AI displacing humans from the workforce FS sector Industry examples of AI-enhanced capabilities
at large scale. Rather that AI augments
the workforce and drastically scales up Retail banking NatWest reduced fraud by 6% as a share of UK
processing capacity and quality. The role of Industry (19% to 13%), including a 90% reduction in
the human workforce will naturally shift to a account opening fraud since 2019 which all contributed
higher level, with a greater focus on design, to reducing operational costs. On the income side they
oversight and exceptions management, achieved a 5x increase in click-through for personalized
as well as having more bandwidth for the lending on customized customer offers. 7
relationship-based, customer-facing roles
where human emotional intelligence is vital. Reduced credit card delinquency by 32% (brighterion
by Mastercard).8
Across financial services (FS) sectors,
Corporate and UK banks have been fully automating the loans
we are seeing the green shoots of AI value
transaction banking underwriting process up to US$100,000 (we have
being realized. Bloomberg was among the
seen up to US$250K).9
first to announce training their own model,
with BloombergGPT providing a means for
JPMorgan Chase developed a GenAI model to analyze
users to query and interact with complex
statements from the U.S. Federal Reserve to determine
financial data using natural language.
the nature of policy signals.10
Citigroup uses GenAI to assess the impact of new US

capital rules.11
Goldman Sachs is working on various projects which will

incorporate GenAI into its business practices. Among
An important the most mature of the projects include writing code in

English-language commands, and being able to generate
documentation.12
point is that Morgan Stanley is using machine learning to identify
we do not see personalized investment ideas and suggest the “Next

Best Action”.13
AI displacing Investment banking (IB)

and capital markets
Projected 27% productivity increase across investment
banks and 27%–35% front office employee
humans from productivity by 2026.14
Insurance Underwriting teams at a specialized insurer experienced
the workforce a 113% productivity increase using generative

AI-supported workflows for underwriting submissions
at large scale. relating to bespoke policies.15
03
2. What impact can AI

have on the bottom
line and how?
04
2. What impact can AI have on

the bottom line and how?
Successful innovators The recent B&CM industry hype around

AI could appear as the latest fad – another
AI will likely now act as the conduit that
accelerates business impact and magnifies
can achieve a 5-15% topic attracting much discussion but not value realization. We look at this more
ultimately leading to sustained operating closely later in this paper.
improvement in cost- margin uplift. Most banks have invested
income ratio over the in strategic innovations in recent years as

cloud, data and digitization technologies
Ultimately, the significance of AI to the
sector will be assessed on the extent
next five years. have advanced. Not all banks have yet that this innovation delivers sustained
achieved material improvement to their operating margin uplift. Here we consider
bottom line from these investments, the “size of the prize” given a typical cost-
particularly where they bolted new systems to-income ratio profile today and expected
and capabilities on to existing technology AI benefit themes.16 We see potential for a
estates, introducing additional cost and 5-7% positive contribution in 2-3 years, and
complexity without decommissioning 10-15% in 5-7 years. This view considers
legacy components. a wide range of banks, and smaller, more
nimble organizations including those with
However, the banks that have learned how currently high cost-income-ratios (CIRs)
to deliver innovation in their organization would find greater opportunity to achieve
will continue to outperform with AI, i.e., the higher end of this 5-15% range of
“the winners will keep on winning”. improvement.17
Successful cloud, data, analytics and
digitization initiatives have provided the
foundational capabilities for AI.
Figure 1. Cost reduction examples
Cost efficiency – examples Typical mid-sized universal bank Income growth – examples
Income = 100
Next generation market analysis/
Trading activities 10 predictive trading algorithms
5–7% uplift on trading income
Fees and
commissions 15
Improved customer retention
Workforce acceleration efficiencies 1–2% uplift fees/commissions
(more from less) across the board Cost = 60
Improved Customer
0–15% total staff cost
Staff Acquisition through
25 hyper-personalised marketing
IT development and
5-10% uplift Interest income and
maintenance acceleration Technology staff fees/commissions
10–20% of IT staff cost 5–10
Credit loss charges 5 Interest income (Net)
Improved credit-risk assessment Tailored loan pricing based
75
leading to fewer impairments Premises and on credit risk assessment
equipment 10 2–3% increase on net
10–-15% saving in impairment charges
interest income
Other administrative/
Improved FinCrime/fraud
accounting
detection reduces litigation/redress 20
charges and fraud losses
Source: © 2024 Deloitte research. For information, contact Deloitte Global.

Note: this is an indicative ‘sizing’ view based on a typical cost/income profile in the industry and our ranged estimates of the potential of AI to improve performance in
specific areas. The examples shown are not exhaustive. Broad-brush costs for implementation/operating costs of AI, and for reduction/redeployment of headcount
are considered, while noting these may vary significantly across different organizations. The cost-income profile shown is informed by third party market data from
Refinitiv, Factica, Statista and selected publicly available Bank Annual Reports as available in Q4 2023.
05
Where will the benefits come from? However, given AI risks and the evolving regulatory
AI, including GenAI, can bring advantages such as: landscape, AI without appropriate human supervision
may not be suitable for:
• Increased efficiency – automate repetitive tasks, freeing
human resources for more complex, creative or customer • Critical, fast-moving operations where timely human
facing engagement. supervision/intervention is not yet feasible.
• Improved accuracy – process vast amounts of data with • Customer/staff facing activities requiring human
greater precision and fewer errors than humans, leading to emotional intelligence (EQ).
more accurate predictions and outcomes.
• Regulatory-sensitive activities.
• Enhanced personalisation – analyze customer preferences
and behaviours to create tailored experiences, improving We see three key modes for achieving value through AI,
customer engagement. all of which combine AI and human strengths:
• Predict trends – make data driven decisions, detecting trends 1. A focus on productivity through personal agents;
and predicting changes in the market.
2. A focus on improving quality and process performance
• Creativity – new possibilities to create new possibilities for through specialist agents; and,
products, services and business models fostering innovation
and growth. 3. Large scale re-imagining of end-to-end processes
using the multi-modal capabilities of AI.
• Cost savings – streamlining operations, reducing errors, and
enabling better decision-making, AI can help save costs and The persona of core “agent modes” in which humans and AI
allocate resources more effectively. interact to implement the operating improvements that can
deliver financial impact. We stress the point that the benefit
• Protection – improving the effectiveness of financial crime in all three modes comes through combining human and AI
and loss prevention capabilities. strengths, not through large scale replacement of humans
with AI. Institutions should develop and strengthen the
• Accessibility – Make the services more accessible human skills to allow for adoption and value realization.
and affordable.
These modes will be leveraged in creating value across the
financial institution.
06
Figure 2. Examples of AI personas
Personal Specialist Transforming

agent focus on agent focus on process focus
productivity improving quality on cost reduction
AI assist AI augmentation AI automation

10-20% potential 20-50% potential 50-80% potential
Executive and specialist roles Those with domain knowledge e.g. Customer facing and support roles
e.g. functional leaders, top levels investment manager, underwriters, e.g. contact centre agents,
of management relationship/account managers central services
Human strengths: Human strengths: Human strengths:

• Emotional intelligence • Relationship management • Problem solving and
• Creativity • Negotiation decision making
• Strategic planning • Domain knowledge • Compassion
• Persuasion and negotiation and experience • AI ethics and regulation
• Motivational leadership • Story-telling and making • AI-Human task management
• Ethical judgement and integrity insights relevant
• Critical thinking Machine strengths:
Machine strengths: • Fraud detection and prevention
• Analyze data and generate Machine strengths: • Data categorization
content • Speed in insight gathering • Quicker processing times
• Schedule meetings • Error checking and • Language translation
• Provide real-time assistance and validation exercises • Voice and text sentiment analysis
suggestions on documents • Trend spotting and
simple graph design
• Trading algorithms
• Predictive analytics
• Routine forecasting
07
Driving down cost through efficiencies and loss prevention

Most banks are currently building AI business cases around cost reduction, and this is no surprise.18 It is easier to get funding approved
for initiatives which drive out cost. The impact tends to be delivered quicker and benefits tend to be more directly attributable to the
investment made. As AI grows in its ability to take on the increasingly sophisticated tasks that previously required human action, the
opportunity grows for banks to perform a wider scope of activities faster and better, doing more with less.
Key cost reduction themes will likely include:
1. Workforce acceleration 2. Engineering transformation

A “marginal gains” approach to deploying Specifically, to benefit bank’s large
many productivity improvements across technology functions, GenAI can already
the human workforce. At the most generate and optimize software code,
Cost
basic level, this will include automation reducing the time to write, while improving efficiency
of repetitive tasks such as data entry quality. As many software engineers examples
and analysis, search and query, draft in banking information technology (IT)
production of many varieties of tend to be relatively inexperienced and
operational content (meeting minutes, requiring oversight from senior engineers,
communications) and summarizing large GenAI “co-pilots” have the ability to
3. Loss avoidance
documentation. This is the type of “text and accelerate production releases and make
Risk management, fraud prevention, cyber,
images” productivity support perhaps most maintenance less onerous.
legal and other brand protection functions
associated with GenAI, particularly among
have high potential for improvement
newer users.
through AI. These functions tend to be
AI applied: a Portugal based improved by speeding up processes,
However, we see that the art of the institution has deployed an expanding scope of processes, and
possible is rapidly expanding, with more AI-powered converter tool that providing wider sets of data inputs to
specialist acceleration use cases including converts software code from legacy improve process performance – all of
data governance and management, COBOL-based systems to their which AI readily supports. Specifically,
data quality and remediation, model target Oracle platform to accelerate AI-enhanced credit risk management
development and analytics. a core platform modernization improvements can result in fewer loan
program. The large language impairments and write-off charges. Fraud
Workforce acceleration will likely require model (LLM) based converter prevention and financial crime (FinCrime)
widespread uplift in workforce skills with AI automatically generates functional processes can be accelerated and
in the same way as staff previously became documentation of the legacy expanded using AI to review a wider set of
proficient in typing, spreadsheets and COBOL code and creates a target input data sets to uncover new insights on
calendar management and other functions metadata schema to accelerate the actors and ultimately reduce losses.
which historically were performed by technical specification and build of
specialist resources only. the new data platform.
AI applied: Legal outcomes
A second use case is the ability for prediction. A Middle-East based
AI applied: various proprietary GenAI to consume millions of lines of bank is trialling a GenAI tool based
GenAI tools are being deployed in legacy code that is undocumented, on past contracts and litigation
compliance teams to summarize and rapidly extract business outcomes. The tool examines the
large sets of documentation issued rules/requirements to accelerate contracts and other documentation
by government and regulators.19 modernization. Deloitte practitioners involved in legal disputes and helps
This rapidly makes the key are already leveraging these the legal team better predict likely
takeaways and major insights capabilities to accelerate client’s outcomes of legal matters, as well
available to compliance teams and transformations and modernize as highlight potential risks in new
business staff in frontline roles. our own products internally.20 contracts.21
08
Growing revenues through new capabilities and improved retention

While more difficult than cost-cutting, players will likely also invest to grow revenue. Revenue growth is a key challenge for banks due to the
relatively limited number of “opportunities to influence”. Consider supermarkets, which have practically limitless opportunities to influence
consumer purchasing behaviour through ranging, discounts, multi-buy offers and more. Unlike supermarket customers, how often do retail
banking customers re-mortgage, change current account or take on a new loan or credit card?
Conversion rate for any sales campaign is a critical metric for AI to improve. When consumers do switch financial products, pricing/rates
are a key factor in the decision, as is trust, and the quality of relationship that the consumer perceives with the bank – influenced by
service level and relevance of interactions and offers. AI can improve all of these factors, while reducing cost to deliver. We see a number
of key revenue-impacting themes:
2. Customer experience and retention

AI-powered digital agents (e.g., chatbots) AI applied: Advanced chatbot.
Income can reduce customer wait times by Bunq, a Netherlands-based
neobank, has recently introduced
growth addressing an increasing range of
complexity of customer requests. While its very own generative AI platform
examples certain customer journeys (e.g., those called Finn. This innovative platform
associated with large transactions, is designed to impress customers
bereavement etc.) must remain as person- with its exceptional ability to
to-person interactions, the improved provide answers to a wide range
1. New capabilities for growth responsiveness of digital customer service of money-related queries. Finn
We see that banks will invest in revenue- agents can improve customer experience features a chat-style text box that
generating capabilities across business and retention rates. Increasingly, the quality allows users to ask questions
lines, including: of AI interaction with humans will improve or seek advice about their bank
as AI technology develops–adjusting the AI account, spending habits, savings,
a. Insight-driven pricing: real-time agent’s behavior according to the behavior/ and other financial matters. The
customization of pricing (e.g., preferential emotions of the customer. platform is capable of combining
lending rates) to make highly competitive data to provide answers that
offers to target customers based on go beyond simple transactions,
enhanced measurement of their AI applied: Service content such as helping users recall past
credit risk. management. A Netherlands-based experiences like “What was the
institution has implemented a name of that Indian restaurant I
b. Hyper-personalized marketing: natural language processing (NLP) visited with a friend in London?”23
improved conversion rates based on chatbot to support front-line staff
insightful identification of individual in delivering a more insightful
prospect and customer/client needs, customer experience. The tool
and highly-tailored communication. enables service staff to query
wide datasets in real-time based
c. Next generation trading algorithms: on live customer requests, rapidly
trading income uplift from enhanced returning relevant responses from
market insight and automated trading product catalogue, account fees,
decisions. terms and conditions, policies
etc. The next phase will enable
customers to interact directly with
AI applied: A UK-based universal the chatbot as a digital agent.
bank has increased click-through
rates on its personal lending offers
by five times, through personalized
offer content and improved target
selection.22
09
3. How will the sector

landscape change and
who will be successful?
10
3. How will the sector landscape change and

who will be successful?
AI is changing the game The competitive landscape will likely be

redrawn, with sector’s probable winners
As mentioned earlier – for leading
institutions, AI is already making the
– who will be the new and losers determined by the speed and paradigm shift from being an instrument
effectiveness with which their AI initiatives of strategy, to a determinant of strategy.
winners? enable evolution of their business
operations, products and services.
Figure 3. How AI is changing the game
Game 1 Game 2 Game 3
• Same processes – • New processes – • New business –

lower cost same business strategy, segments,
• Step change in • Transform products, service,
efficiency and customer experiences
productivity (cloud, experience, • Distinctive
automation) personalise definition of
products/services purpose /
(digital, data, AI) contribution to
society
• Expanding the Art
of the Possible (AI/
GenAI)
High
Changing
the game...
Game 3
Intelligent
banking
Value potential
Game 2
Digital
banking
Game 1
Classic
banking But how to jump
from Game 1 to
2 to 3?
Low Time High
11
So, who will be successful? How? Execution is critical

Banks expected to capture the biggest Key technical foundations The players which have realized benefits
benefits from prior waves of technology- • Cloud, where done well, has delivered from prior technical innovations have
enabled innovation (e.g., cloud, digital, data) readily-scalable computing power learned and refined the delivery methods
will continue to outperform in their value and accessible data provisioning, that work in their organization. Typically,
creation from AI. This is because leaders in that abstracted data away from the these have included consideration of:
innovation have already invested in the key complexity of legacy architectures while
organizational enhancements, including reducing total cost ownership of the • Governance – putting in place sufficient
culture, governance, data management IT estate. It also forced banks to learn oversight to adequately assess and
and agile delivery methods, needed to how to assess and manage the risks mitigate the spectrum of risks, without
capitalize on the AI opportunity. associated with introducing third-party unduly constraining delivery;
dependencies to the infrastructure
In many ways, substantial prior supporting core business processes. • Culture – benefits are well
investment in the innovations (cloud, data communicated, business function owners
management etc.) mentioned above has • Automation put in place the governance expect to embrace emerging technology
prepared the ground for value creation and risk management capabilities to to improve process performance;
from AI. All of these investments required oversee automated operations.
considerable capital expenditure that • Idea to value – strong processes
has constrained the bottom-line benefits • Data governance may have been are embedded to generate ideas for
realized to date. However, as above, implemented initially for compliance value delivery from innovation, assess
the organizations that have successfully purposes but has established the feasibility and investment case, rapidly
invested in these ambitious infrastructural organizational accountabilities, policies, deliver the best ideas into production
changes will find AI to be the conduit that quality improvement methods and and scale;
now accelerates the unlocking of value. understanding of organizational data
assets to provide trusted datasets as • Talent – hiring and learning/development
inputs to AI use cases. approaches that build adequate skills and
capacity; and,
• Digital banking has evolved customer
expectations to be more comfortable • Partnerships – engaging with the wider
with self-service, real-time, insight-driven market ecosystem, forming partnerships
and reduce reliance on bank staff for with technology and service providers
many interactions, while streamlining best placed to assist delivery.
key front-to-back processes e.g., client
onboarding, loan fulfilment.
12
Realizing the value from AI will take

more than simply enabling the
The continuous upskilling of teams who
use these new tools to do more is not The continuous
technology. In recent history there a one-time effort, it should be built into
have been great expectations that
technology transformation will drive
the talent model and measured. Banks
who simply implement AI and GenAI to
upskilling of
significant efficiency gains only to deliver
underwhelming results. Global chief
augment existing processes will likely
not see the full value realization and teams who use
technology officer of Dell Technologies Inc, could in fact only see increased costs.
captures the frustrations of many senior
executives with the sustained investment
Banks who leverage AI and GenAI to
support continuous transformation and
these new tools
required: “I must’ve had ten conversations
last week where CIOs were bemoaning
improvement can take the foundational
investments already made (e.g., cloud and to do more is
that they had run out of money or blown data) and unlock further value.
their [cloud] budget off.24 ”Why could it be
different this time? The past investments Considering these points, the FinTech
not a one-time
(e.g., cloud, automation) have been parts
of a solution but ultimately have not yet
subsector is likely to move quickest, due to
distinct execution advantages. Namely: effort, it should
delivered transformational bottom-line
value. In the case of cloud, organizations
may have built the new capabilities but not
• The relative simplicity of their current
operating models (considering
be built into the
yet switched off what these cloud-based
solutions were intended to replace. In
products, processes, technology, data
and organization) makes them less talent model
the case of automation, it was possible to encumbered by the constraints of legacy
automate parts of a process with great
precision but the technology struggled
systems and processes. They still have
the flexibility to jump straight to newly-
and measured.
with inferencing and being intuitive; it was conceived processes without lengthy
a brittle solution in areas that required re-engineering of legacy.
elasticity to be effective.
• They typically have a culture tilted to
AI is already interacting with the workforce more rapid growth and innovation – their
in a more natural way and opens the doors greater risk appetite means they will be
for entirely different processes. Solutions willing to push AI capability to customers
for these processes can now be developed and into production processes sooner.
not as 1’s and 0’s but rather with natural But there are risks associated with doing
language providing great flexibility and this, before having the appropriate
speed to solution. guardrails and risk infrastructure in place.
Therefore, true value to banks will be

delivered when costly and long duration
processes are reconceived. As banks
evolve in maturity with AI and GenAI they
will begin to give front line employees
increasing autonomy and improved tooling
that will enable increasing revenue
(see “insight-driven pricing”) while also
reducing non-value add work (e.g., data
entry). But once that tooling is in place and
banks begin to reconceive processes there
must be a focus to continue to redeploy
staff to higher value roles.
13
4. What is special about

generative AI and where is
this technology heading?
14
4. What is special about generative AI

and where is this technology heading?
GenAI is a branch of AI currently attracting GenAI is about more than just text:
much attention, as it allows for the Gen AI is capable of working with multiple
generation of increasingly sophisticated “modalities” of content, with the ability to
content (e.g., text, code, audio, images, process one modality as input and generate
videos, processes) based on algorithms that another as output (not all combinations
imitate existing content, using statistical shown in figure 4 are currently possible).
predictions learned from large sources. Gen AI is able to produce sophisticated
content output including software code,
The fast-improving apparent quality of PowerPoint presentations and three
this content suggests that GenAI can dimensional (3D) models.
play a large role in business functions
traditionally considered to require
solely human intelligence.
What is different about GenAI

and why all the excitement?
GenAI is predicted to be the start
GenAI rapidly generates sophisticated
content, based on vast bodies of source of a new era for AI. The technology
information, designed to imitate what
a skilled human being could produce.
This could be for example summarizing
will continue to evolve with focus
large volumes of documentation, writing
an opinion piece, developing software
on multi-modal communication
code, producing images/video to a
given specification, preparing a sales and intelligence built into human
presentation or defining rules to
measure data quality. interactions.
Figure 4. AI modalities
Text Text
Code Code
Audio Audio
Image Image
Video Video
3D/Specialized Read more about the next six 3D/Specialized

modalities in our recent publication
Generative AI Dossier.
15
The increasing sophistication and GenAI is already being used: According

What are the limitations? to Deloitte UK’s Digital Consumer Trends
apparent quality of this generated
2023 report, around a third of users
output is compelling, and points • It is just imitating human output (32%) claimed to have used GenAI for
to an opportunity to augment by predicting statistically what work.26 This is 8% of the UK population, or
organizations’ human workforce’s human generated content would approximately 4 million people. Given the
look like based on vast volumes of lack of corporate policy and governance,
at large scale.
previously published content. it is reasonable to assume that much of this
use was unsanctioned, and without proper
• It cannot “sanity-check” or education employees may have been at
What is it good at? challenge the output (although risk of sharing confidential information and
other forms of AI can do this). failing to recognize hallucination and bias.
• Tasks that a human would do
far slower, generating content • It can construct statements Where will GenAI go from here?
based on research, or vast that may appear plausible but GenAI will continue to evolve. We see
amounts of information. are simply untrue (known as three themes of increasing maturity:
hallucinations).
• Spotting trends or anomalies
1. General intelligence: the mimicry
across large datasets at speed, • It can be biased, amplifying the of human thinking will likely mature
that a human may miss. inherent biases in source content. to include reasoning, making new
inferences and predictions based on
• Augmenting human teams to • It depends heavily on quality of complex inputs. It will start to form
accelerate output of a function source content. perspectives and views that challenge
or expand coverage.
human thinking. Currently GenAI
• Large computing power required focusses on distilling, summarizing
• Producing specialist content on leading to potentially greater and producing human-like content.
demand – e.g., images, video, carbon emissions associated 2. Expanded modalities: the available
3D models. with AI.25 combinations of input/output
modalities will increase, with greater
• Increasing customization, i.e., • No EQ – does not adapt interaction sophistication in the production of
with GenAI effective scale, with human beings based on specialist output such as financial
supports “segment of one” emotions of the human user. engineering constructs, systems
in customer relationship
architecture design, audit opinions
management.
and risk control assessments.
In future developments, we see AI evolving 3. Artificial EQ: interactions will become
to display improved apparent EQ with more human-like, developing the
humans, responding appropriately to facial ability to change tone and respond
and speech cues to become increasingly appropriately to human emotions.
suitable in customer service agent roles. This will be particularly impactful,
making AI suitable for an increasing
number of human-facing use cases.
16
5. How will banks embed AI

across the value chain?
17
5. H
ow will banks embed AI across the
value chain?
We see value creation being achieved via the build of an inventory of combined human/AI agents, deployed right across the value chain.
Initially, these agents would improve provide marginal gains in workforce productivity, and performance improvement of specific
processes through quality improvement and/or automation. The operating model will look like today, but will be faster, better, cheaper, and
less risky. Depending on organizational priorities, i.e., achieving cost efficiencies, risk reduction or focussing on the customer experience
and growth agenda, different functions across the Bank will have differences in potential. Below we set out focus areas for different
organizational priorities against a simple universal banking value chain.
Figure 5. Value chain focus areas according to organizational priorities
Strategy and Sales/trading Customer/client Operations Risk and

proposition and marketing experience and service compliance
Cost
Customer
Growth
Risk
Enabling and control functions
Technology Data and analytics Finance HR Legal
Cost
Customer
Growth
Risk
Key: Lower potential High potential Highest potential
18
Marginal gains: Impact of combined AI and human capability across the value chain
We expect that whichever components of the value chain are prioritized, results will come from building and implementing enhanced and/
or new capabilities that combine human and AI strengths. Deloitte has defined an initial priority set of over 50 banking-specific use cases,
many of which we are already working with clients to implement, which will likely have widespread impact across the entire value chain.
A selection of examples, including the business challenges they address, are described below. We use the word “agent” to indicate a new
capability that combines AI and human inputs in its operation.
Value chain Typical business challenges AI opportunities

component
Strategy and • Increased competition • Research agents to summarize and perform

proposition trend analysis of market, customer, channels and
• Specialist resource constraints
products to support strategy and proposition
• Internal siloes across propositions development.
• Lack of data to inform strategy and proposition

development
Sales/trading and • Evolving customer/client landscape and • Hyper-personalization communication and

marketing expectations personalization agents driving personalized
messaging and marketing campaigns based on
• Identifying individual needs and tailoring products/
market available data, interaction data (voice,
services accordingly
digital), etc.
• Omni-channel marketing and personalized content
• Creative content creation (multi-modal) to
• Financial market volatility accelerate delivery of the marketing content and
aligned to brand, legal and compliance policies.
Customer/client • Lack of customer/client-centricity • Virtual agent scaling the operations centre

experience automating end-to-end client experience.
• Capturing and managing customer/client feedback
throughout touchpoints • Assistant for channel engagement (RM, call-
center, branches and email) improving efficiency,
• Establishing seamless omni-channel presence
driving improved customer experience.
integrating physical and digital experiences
• Availability of insights to customer/client profile for

service agents/risk management (RMs)
Operations and • Straight-through-processing and real-time • Client/customer on-boarding by gathering

service capability information, assessing trustworthiness and
seamlessly onboarding them onto new products.
• “One pattern for all” model in legacy systems/
processes restrict personalization • Customer service managing complex requests
seamlessly, helping customers engage with their
• Poor data management restricts quality insights
bank and providing them with useful responses.
and impairs service levels
• Fees collection and recovery services for
• Fragmented onboarding and service journeys due
early detection and helping draft appropriate
to organizational/ technical silos
communications.
19
Risk and compliance • Complex and evolving regulatory requirements, • Regulatory compliance by detecting, alerting
particularly for global institutions/organizations. and generating triggers for the customers.
The summarization for compliance reporting.
• Emerging risks due to rapid technological change –
particularly AI and cyber. • Financial crime support activities like, narrative
generation. Adverse media screening and other
• Poor data management hampers the ability to
potential fraudulent threats.
monitor, measure and manage risks.
• Cyber detection services to look for any
• Balancing risk appetite and growth ambitions.
anomalies and detect patterns.
Technology • Leveraging technology to improve operational • Delivery of engineering and operations

efficiency, enhance customer experience, support to drive quick delivery of products/
maintain compliance and ensure security – solutions to market.
all while remaining agile and competitive
in an evolving financial landscape.
• Legacy systems with poor real-time capability.
• Skills and capability challenges in retaining

knowledge of legacy/proprietary systems
and new technology.
Data and analytics • Banks are data-intensive organizations and • Accelerated insight production where business
yet data required for critical business processes can use where natural language to query complex
can be fragmented, poorly understood and datasets without reliance on technical data staff.
not of fit-for-purpose quality.
• Greater coverage of data governance
• Increasingly stringent regulatory requirements standards, where GenAI accelerates data profiling,
increase the load on business and chief data definition and quality improvement.
office teams.
• Increasing dependency on data and analytics to

deliver business growth objectives.
Other enabling and • Complex financial regulation. • Focus on functions like, HR, finance, legal, etc.
control functions to develop specialist agents to integrate into
• Costly legal litigation and disputes.
business processes providing summarization,
question and answer (Q&A) chatbot or
writing documents.
Deloitte has defined an initial priority set of over

50 banking-specific use cases, many of which we
are already working with clients to implement.
20
Depending on organizational Figures 6, 7 and 8 are based on Deloitte

priorities and ambition, banks should research and analysis of the value chain
look in depth for opportunity in across banking clients. The research also
specific functions identified multiple “strings of pearls”, i.e.,
If cost efficiency is the priority, then chains of related processes which can be
banks should look at the largest functions, re-imagined end-to-end to help deliver
and those where the nature of processes substantial additional value. Please contact
mean that AI can deliver substantial the authors to discuss the applicability to
efficiency or cost avoidance benefits. your organization.
Our analysis of level 2 processes indicates
that the technology organisation, risk and
compliance and customer experience
offer particularly high potential for cost
savings. In Figure 6 we set out a view of the
highest potential level 2 processes for cost
efficiency, organized by the relative scale of
the function and the level of cost efficiency
and/or avoidance expected.
Figure 6. Cost efficiency agenda – bank functions with greatest potential benefits from AI
Larger
Fraud & Financial
Crime Screening Non-Financial Credit Risk Regulatory
Credit Screening Risk Management Management Compliance Transaction
& Underwriting Monitoring
Customer Customer Support
Relationship (e.g. Branch, Phone,
Management Social Media)
Digital Support
Collections & Litigation Financial Crime MLRO
Relative Improvement Through AI
Recoveries Data Management

& Governance Data
Data analytics
Engineering Account
IT Incident Management Application
IT Security
Management Development
& Recovery
& Maintenance
Regulatory Reporting IT Service
Management
Market Risk IT Service Desk
Management IT Engineering
Card
Payments Management
Customer Management (In/Out) Campaign
Platform
Communications Planning &
Management
Management
Digital
Workplace
Services
Customer
Journey Management
Larger bubble size indicates
greater cost/complexity to
achieve potential improvement.
Smaller Relative Scale of Function Larger
21
Similarly, where risk reduction is prioritized, the largest opportunities appear within risk and compliance, data management, service
operations and legal functions. Note we see correlation in priority focus areas for the cost efficiency and risk reduction agendas, for example
in transaction monitoring, financial crime and regulatory compliance. This is because risk events in these areas have high cost implications.
Figure 7. Risk reduction agenda – bank functions with greatest potential from AI
Credit Screening Non-Financial Credit Risk Transaction

& Underwriting Risk Management Financial Crime MLRO
Fraud & Financial Management Monitoring
Crime Screening
Regulatory
Data Management Market Risk Compliance
Enterprise AI
Data Analytics & Governance Management
IT Security
& Recovery Customer Relationship
Onboarding Management Litigation
Exception Handling
Account Management
Risk Model IT Incident
Analytics Collections & Recoveries Management
Risk & Strategy Regulatory Reporting
Governance QA & Testing
IT Service
Document Management
DevSecOps
& Contract
Management Data
Platform Management
Engineering
Asset Liability Liquidity Management
Management Payments Larger bubble size indicates greater
Reconcilliations Management cost/complexity to achieve potential
(In/Out) improvement.
Relative Scale of Function
Finally, organizations prioritizing customer/client experience and growth agenda will likely naturally focus further up in the front office
sales, marketing and service functions, and in tailoring product/service features and pricing. Interestingly, we see credit risk management
as a key enabler of growth, where greater precision in credit-risk assessment enables insight-driven pricing of lending to secure the most
desired customer segments.
Figure 8. Customer and growth agenda – bank functions with greatest potential from AI
Feature Product Launch

Customisation Complaints Management
Campaign
Planning &
Customer Execution
Relationship
Marketing Management
Strategy Customer Analytics & Insight
Content
Customer Development Digital Support (e.g. Customer Support
Behaviour & Production Chatbots, FAQs, Live Chat) (e.g. Branch, Phone,
Data Social Media)
Leads Management Customer Communications
Product Design
Product Management Credit Risk
Data Analytics Management
Machine
Enterprise AI
Learning Loyalty & Rewards Strategy
Next Best Action
Social Media Management
Customer
Proposition Development Customer ON-Boarding Journey Management
& Open Banking Connectivity
SEO & Paid Research Larger bubble size

Publishing & Distribution indicates greater
cost/complexity to
Marketing Asset Management
Commercial Strategy Account Management achieve potential
Marketing Positioning & Strategy improvement.
Relative Scale of Function

22
6. What risks must be

managed and how?
23
6. What risks must be managed and how?
Key risks
In many ways, deployment of AI to scale up
Cyber risk: The risk of irresponsible
application of AI is associated with In many ways,
operations raises similar enterprise risks the various use cases that will likely
as outsourcing to a third party. The role
of bank staff changes to one of definition,
get contemplated (e.g., using LLMs
for heightened automated cyber
deployment
governance and accountability, rather than
direct execution of the process in question.
threats). Additionally, where AI starts
to form substantial components of of AI to scale
The bank is accepting a certain level of an organization’s operations, lack of
increased risk (requiring mitigating controls)
for the benefits that come from increased
transparency into precisely how the AI is
functioning may leave the organization
up operations
processing capacity and lower cost. unable to design appropriate controls and
cyber protections. raises similar
Key risks that should be addressed include
the following: Sovereignty restrictions: The risk of
sovereignty relates to the expectation
enterprise risks
Misuse of AI: incidents involving malicious
actors are a real threat. For example,
that AI models trained on certain data
sets are subject to sovereignty/residency as outsourcing
where deepfake techniques have been regulations and will be required to be
used to successfully imitate a customer,
gaining access to their account commit
run only on data centers within that
jurisdiction. Additionally, organizations
to a third party.
fraud. Deepfake in this case is an artificial, with global processes, for example trading
AI-generated video or sound recording operations, will face complex challenges in
designed to convincingly appear authentic.27 satisfying divergent local regulations across
the jurisdictions in which they operate.
Environmental impacts: Greater use of
AI will naturally require greater computing Future regulation: The risk of lack of
power, in turn leading to greater energy certifications concerns the possibility that
consumption in the data centers. This is a LLMs may face future regulation as they are
competing factor against industry net zero increasingly used for insights, advice, etc.
commitments and climate change impact (e.g., similar to how lawyers must past the
disclosure and reporting requirements. Bar exam).
Amplification of biases: Underlying Autonomous vs. human intervention:

datasets contain inherent biases that will The risk of safe usage is associated with
be amplified once the models are trained how and where LLMs are used (e.g., using
on them, potentially exacerbating any LLMs to generate autonomous action for
discriminations based on gender, race and machinery in a factory floor).
other characteristics. This could expose
banks to litigation and/or regulatory Fight for talent: Across the industry,
consequences. the need for skilled science, technology,
engineering, and mathematics (STEM)
capability, including data science, is on the
rise. Organizations will likely struggle to
retain talent, and will need to put in place
upskilling and recruitment initiatives.
24
Building trust
Figure 9. Deloitte’s trustworthy AI framework
Fair/impartial Robust/reliable
Generative AI can embed Current LLMs have significant
discriminatory or harmful biases accuracy and performance issues
and unable to specify uncertainty
AI governance
Regulatory
compliance
Transparent/explainable Trustworthy Privacy

Generative AI does not produce AI TM Generative AI systems tat are
explanations out of the box trained on all openly available
data may have significant risks of
copyright infringement and/or
privacy concerns
Responsible/accountable Safe/secure
There should be a human-in-the-loop Significant potential for harm in
with sufficient understanding of
AI carries significant
spreading misinformation;
generative AI to challenge and risks and ethical environmental harm; potential
override its outputs considerations malicious actors
Deloitte’s Trustworthy AI framework • Enterprise data sets: Train LLMs with • Large language model operations
sets out guardrails to address key risks. data sets that are governed within an (LLMOps): Form dedicated team(s)
These guardrails include: enterprise rather than the wider internet. focused on operating, managing, and
governing the models to prevent drift
• Secure environments: Train LLMs in • Audit trail: Persist the data that LLMs over time and for rooting out biases.
secure environments in a data center or are trained on to trace the data, map the
on cloud to reduce probability of leakage lineage, and have an audit trail of what
of company information. type of data was used.
• Restricted usage: Restrict initial usages • Trust but verify: Keep humans involved
of generative AI to increase accuracy of throughout the process to assist in
inferences; then scale once there is a validating and verifying the generated
growing sense of comfort. output and to monitor the AI’s accuracy.
25
7. What are the key

considerations for safe
and effective execution?
Principles for success
26
7. What are the key considerations for safe and

effective execution? Principles for success
Executing AI with intent

Banks should execute with a clear vision, that addresses the following concerns for the specific shape of their enterprise.
Figure 10. Business concerns to address
Strategy and Partnerships Platform data Use case model

prioritization and alliances and integration selection
How will you prioritize Who are your partners How will you integrate For prioritized use cases,
and plan to integrate and how will you large language models explore the benefits on
Generative AI use cases orchestrate the growing (LLMs) into your enterprise customer and colleagues,
& capabilities into your risks, costs - run
partner ecosystem? business and data
enterprise business experiments to reduce
architecture?
uncertainty.
strategy?
Operating
Trustworthy AI Legal
model
What organizational How will you ensure How can you mitigate IP
structures, capabilities, that Gen AI usage is infringement and craft
skillsets and processes secure, transparent and supplier contracts that
do you need to support responsible? clearly define liability of
Gen AI? GenAI misuse?
Managing the input data

AI output can only be as good as the quality
Human-centred design
Humans are critical to the design, Humans are
of data and content available as input. development and successful operation
Additionally, AI outputs, particularly those
that drive decisions impacting customers,
of AI. Additionally, human staff should
retain accountability for AI operations,
critical to
should be transparent, fair, lawful, ethical
and secure. Strong governance is critical to
being able to oversee AI processes and
take action where necessary to manage the design,
support traceability, reduce hallucinations unwanted behaviours or outcomes. This
and manage bias. This is only possible
if enterprise source data assets are
means AI functionality should support
human interactions that are transparent,
development
understood, owned and managed. explainable and intuitive.
and successful
Most banks have established data
governance capabilities that provide a
level of control over data flows, ownership,
operation of AI.
quality, definitions and modelling. These
governance capabilities have traditionally
been applied to granular data attributes
(particularly where needed for the Basel
Committee on Banking Supervision’s
standard number 239 [BCSB239]
requirements),28 however models should
evolve to govern far larger datasets.
27
8. How to get started, scale

and drive adoption
28
8. How to get started, scale and drive adoption
Collaborating with Achieving sustained value requires

organizational evolution
In executing the AI strategy,
key building blocks need to be
several leading banking As mentioned in earlier sections, we see established early on
sustained value being delivered through We believe the following are essential
clients on their journey new capabilities that combine human and building blocks for standing up an effective
to transform business AI strengths. Accordingly, organizational

development to manage AI is as important
AI change agenda. These building blocks
span vision, governance, people, process
performance through as the technology itself and requires a set and technology.
of steps forward.
AI. Our experience 1. Strategic blueprint: Define the
highlights the criticality • Mindset evolution: Move beyond

the endless cycle of near-term proof of
business ambition with clear short-
and long-term goals and investment
of organizational and concepts, and place long-term bets on AI commitments.
in key areas.
cultural evolution 2. Operating structure: Build a model to
alongside a set • Leadership evolution: Set leadership

goals against measurable AI targets
source, prioritize, govern and develop
AI/GenAI capabilities.
of technological and value, to drive evangelism and
accountability. 3. Value realization: Stand up a value
and programmatic tracking mechanism to measure
must-haves. • Investment evolution: Treat AI

investments as core enablers of
outcomes and optimize investments.
enterprise business strategies and 4. Capabilities and technology: Create

not as experimental investments. a technology and partner roadmap to
enable AI/GenAI requirements.
• Cultural evolution: AI should be seen
as a skill that all employees will need 5. Ways of working: Build enterprise AI
to embed in their ways of working fluency and optimize the delivery and
(particularly GenAI). usage operating model to maximize
adoption.
• Execution evolution: Move beyond
the front and back-office method and 6. Responsible AI: Establish ground rules
adopt a “two in the box” approach where and promote accountable and ethical
business and IT are set shared goals. use of AI.
29
We target an 18-month journey to driving value and scale

Figure 11. Timeline for driving value and scale
1 2 3 4
Strategize and mobilize Foundational bets Iteratively scale Institutionalize AI
(3 months) (6 months) (6 months) (ongoing)
• Develop an AI • Deliver quick wins • Scale MVPs and • Deliver new use cases
strategic blueprint from efficiency plays embed AI into to complete end to
• Mobilize stakeholders • Develop strategic business workflows end transformation
• Identify high-value partner ecosystem • Measure outcomes of value streams
transformation • Develop enterprise from adoption • Develop enterprise
opportunities controls for • Build AI fluency and AI assets and
responsible adoption upskill workforce capabilities to
Result: AI North Stars and unlock efficiencies
governance framed, key use Result: MVPs developed; key Result: AI savvy workforce,
cases identified technology and governance initial value realization Result: Scaled business
decisions made transformation
How to get started?

Our tested action plan:
1 2 3 4 5
Establish a Align on priority Identify no- Create minimum Scale solutions
leadership mandate: focus areas: regrets bets: viable governance: and realize value:
Empower a leader(s) Prioritize two-three Activate business Establish a governance Scale AI solutions and
with a mandate opportunity areas to units and IT/digital function that can capabilities across
to own and drive and serve as overarching to identify initial manage AI risks, business value streams
enterprise north stars (e.g., “no-regrets bets” investments and and measure outcomes
AI/GenAI agenda product launch, that align to progress, while to progressively reach
credit risk) priority areas. encouraging peak value realization.
innovation.
30
9. Contacts
31
9. Contacts
For further information, or to discuss your AI challenges, please contact us.
Zhongbin You
FSI Solutions Offering Leader, Deloitte Consulting China
AI Institute Co-Managing Partner, Deloitte China
Tel: +86 21 2316 6172
Email: [email protected]
32
Endnotes
1 “Autopilot and Full Self-Driving 16 Sourced from Refinitiv, financial

Capability”. Tesla. March 2024. statements of global 25 major
universal banks.
2 “AI in pharma and life sciences”.
Deloitte. 2024. 17 Sourced from “Refinitiv”, “Factiva”,
“Statista” selected Bank Annual
3 “Facial recognition technology”.
Reports as available in Q4 2023.;
Metropolitan Police. 2024.
“Deloitte AI Institute”. 2024.;
4 “The EU AI Act” Deloitte Insights 2024. “Gen AI Dossier”. 2024.
5 “Jamie Dimon’s JPMorgan and other 18 ethink: How banks can drive
R
North American banks are so far down cost and protect profit”.
ahead in the AI race that one indexing Avanade. 2020.
boss says it’s now a question of ‘Can
19 “How generative AI can help banks
others catch up?’” Fortune. April 2024.
manage risk and compliance”.
6
“JPMorgan to invest $1bn or more McKinsey & Company. March 2024.
a year in AI, Daniel Pinto says”.
20 “Modernization and Code Generation”.
Nikkei Asia. August 2023.
IBM. 2024.
7 “Data strategy roundtable”.
21 “Middle East GenAI spotlight”.
NatWest Group. June 2022.
PwC. 2024.
8 “Credit risk datasheet”. Brighterion.
22
Data strategy roundtable”.
2020.
NatWest Group. June 2022.
9 “2024 banking and capital markets
23 “Bunq launches GenAI chatbot.”
outlook”. Deloitte Insights.
Finextra. December 2023.
October 2023.
24 “CIOs Still Waiting for Cloud
10 “JPMorgan Creates AI Model to
Investments to Pay Off”.
Analyze 25 Years of Fed Speeches”.
Wall Street Journal. September 2022.
Bloomberg UK. April 2023.
25 “ The surge in AI is straining the
11 “Citi Used Generative AI to Read
U.S. power grid”. Fast Company.
1,089 Pages of New Capital Rules”.
March 2024.
BNN Bloomberg. October 2023.
26 “Digital Consumer Trends 2023”.
12 “ Goldman Sachs developing dozen
Deloitte. August 2023.
generative AI projects – exec”.
Reuters. November 2023. 27 “Deepfake”. Cambridge Dictionary. 2024.
13 “The Future Of Work Now: Morgan 28 “Basel Committee on Banking

Stanley’s Financial Advisors And Supervision. Principles for effective risk
The Next Best Action System”. data aggregation and risk reporting”.
Forbes. May 2020. Bank for International Settlements.
January 2013.
14 “Unleashing a new era of productivity
in investment banking through
the power of generative AI”.
Deloitte Insights. July 2023.
15 “Case study – Markel uses Cytora and

achieves +100% productivity uplift to
fuel growth”. Cytora. September 2023.
33
About Deloitte
Deloitte China provides integrated professional services, with our long-term
commitment to be a leading contributor to China’s reform, opening-up and
economic development. We are a globally connected firm with deep roots locally,
owned by our partners in China. With over 20,000 professionals across 31
Chinese cities, we provide our clients with a one-stop shop offering world-leading
audit, tax and consulting services.
We serve with integrity, uphold quality and strive to innovate. With our
professional excellence, insight across industries, and intelligent technology
solutions, we help clients and partners from many sectors seize opportunities,
tackle challenges and attain world-class, high-quality development goals.
The Deloitte brand originated in 1845, and its name in Chinese (德勤) denotes
integrity, diligence and excellence. Deloitte's global professional network of
member firms now spans more than 150 countries and territories. Through our
mission to make an impact that matters, we help reinforce public trust in capital
markets, enable clients to transform and thrive, empower talents to be future-
ready, and lead the way toward a stronger economy, a more equitable society
and a sustainable world.
Deloitte refers to one or more of Deloitte Touche Tohmatsu Limited (“DTTL”),

its global network of member firms, and their related entities (collectively, the
“Deloitte organization”). DTTL (also referred to as “Deloitte Global”) and each
of its member firms and related entities are legally separate and independent
entities, which cannot obligate or bind each other in respect of third parties.
DTTL and each DTTL member firm and related entity is liable only for its own acts
and omissions, and not those of each other. DTTL does not provide services to
clients. Please see http://www.deloitte.com/about to learn more.
Deloitte Asia Pacific Limited is a company limited by guarantee and a member

firm of DTTL. Members of Deloitte Asia Pacific Limited and their related entities,
each of which is a separate and independent legal entity, provide services from
more than 100 cities across the region, including Auckland, Bangkok, Beijing,
Bengaluru, Hanoi, Hong Kong, Jakarta, Kuala Lumpur, Manila, Melbourne,
Mumbai, New Delhi, Osaka, Seoul, Shanghai, Singapore, Sydney, Taipei and Tokyo.
This communication contains general information only, and none of DTTL, its
global network of member firms or their related entities is, by means of this
communication, rendering professional advice or services. Before making any
decision or taking any action that may affect your finances or your business, you
should consult a qualified professional adviser.
No representations, warranties or undertakings (express or implied) are given

as to the accuracy or completeness of the information in this communication,
and none of DTTL, its member firms, related entities, employees or agents shall
be liable or responsible for any loss or damage whatsoever arising directly or
indirectly in connection with any person relying on this communication.
© 2024. For information, please contact Deloitte China.

Designed by CoRe Creative Services. RITM1806717

Ai 1

Uploaded by

Copyright:

Available Formats

Ai 1

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Ai 1

Uploaded by

Copyright:

Available Formats

Changing the game:

the impact of artificial

1. Where are banks

1. Where are banks in the AI journey?

Artificial intelligence Artificial Intelligence (AI) is already here and

markets sector’s autonomous vehicles operating on public

years. The journey process data and automate workflows

AI now allows banks to tackle

An important point is that we do not see

Citigroup uses GenAI to assess the impact of new US

Goldman Sachs is working on various projects which will

An important the most mature of the projects include writing code in

we do not see personalized investment ideas and suggest the “Next

AI displacing Investment banking (IB)

humans from productivity by 2026.14

Insurance Underwriting teams at a specialized insurer experienced

the workforce a 113% productivity increase using generative

at large scale. relating to bespoke policies.15

2. What impact can AI

2. What impact can AI have on

Successful innovators The recent B&CM industry hype around

income ratio over the in strategic innovations in recent years as

Figure 1. Cost reduction examples

Source: © 2024 Deloitte research. For information, contact Deloitte Global.

Figure 2. Examples of AI personas

Personal Specialist Transforming

AI assist AI augmentation AI automation

Human strengths: Human strengths: Human strengths:

Source: © 2024 Deloitte research. For information, contact Deloitte Global.

Driving down cost through efficiencies and loss prevention

Key cost reduction themes will likely include:

1. Workforce acceleration 2. Engineering transformation

Growing revenues through new capabilities and improved retention

2. Customer experience and retention

3. How will the sector

3. How will the sector landscape change and

AI is changing the game The competitive landscape will likely be

Figure 3. How AI is changing the game

Game 1 Game 2 Game 3

• Same processes – • New processes – • New business –

Low Time High

Source: © 2024 Deloitte research. For information, contact Deloitte Global.

So, who will be successful? How? Execution is critical

Realizing the value from AI will take

Therefore, true value to banks will be

4. What is special about

4. What is special about generative AI

What is different about GenAI

3D/Specialized Read more about the next six 3D/Specialized

Source: © 2024 Deloitte research. For information, contact Deloitte Global.

The increasing sophistication and GenAI is already being used: According

5. How will banks embed AI

Figure 5. Value chain focus areas according to organizational priorities

Strategy and Sales/trading Customer/client Operations Risk and

Enabling and control functions

Technology Data and analytics Finance HR Legal

Key: Lower potential High potential Highest potential

Source: © 2024 Deloitte research. For information, contact Deloitte Global.

Value chain Typical business challenges AI opportunities

1. Where are banks

2. What impact can AI

2. What impact can AI have on

3. How will the sector

3. How will the sector landscape change and

4. What is special about

4. What is special about generative AI

5. How will banks embed AI

6. What risks must be

6. What risks must be managed and how?

7. What are the key

7. What are the key considerations for safe and

8. How to get started, scale

8. How to get started, scale and drive adoption

1 “Autopilot and Full Self-Driving 16 Sourced from Refinitiv, financial

13 “The Future Of Work Now: Morgan 28 “Basel Committee on Banking

15 “Case study – Markel uses Cytora and