FINANCIAL INSTITUTIONS

AI IN FI REPORT 2025

For Banks, the AI

Reckoning Is Here
May 2025
By Saurabh Tripathi, Stiene Riemer, Matteo Coppola, Kirsten Rulf, Jürgen Rogg,
Christian Schmid, Michael Strauß
Contents
03 Introduction

04 The Next Five Years Will Define the Next 30

07 Rethink, Retool, and Reclaim Advantage

11 Act Now
Introduction
AI is moving fast—but Yet BCG analysis finds that only a quarter of financial
institutions are using AI to reinforce their competitive position.

most banks aren’t keeping The rest are experimenting at the margins—understandable,
but insufficient. Winning in the coming era will require far

up where it matters.
more than isolated pilots or cautious upgrades. It means
anchoring AI strategy in business strategy, systematically
prioritizing high-ROI use cases, embedding clear performance
metrics, and mobilizing capital and CEO leadership decisively
Predictive AI has been reshaping financial services for toward scalable impact—with tight alignment between
years, pushing banks to fend off digital competitors and investment and results.
rethink their own operations. But the arrival of GenAI is
an inflection point—accelerating AI’s impact, raising AI presents a strategic fault line. The banking leaders
new strategic questions, and amplifying both opportunities that are acting now—defining where they will compete and
and risks. Agentic AI—systems that can act how they will win—have already begun to shape the
autonomously within set parameters—propels this shift industry’s future. The rest will be shaped by it.
even further by moving AI from analysis into execution.

Together, predictive, generative, and agentic AI are

redefining the banking business—and eroding pillars that
have long underpinned banks’ advantage. The next several Predictive, generative, and
years will usher in new and differentiating capabilities for
organizations that seize and scale them. agentic AI are eroding pillars
that have long underpinned
banks’ advantage.

BOSTON CONSULTING GROUP FOR BANKS, THE AI RECKONING IS HERE 3

The Next Five Years Will Define
the Next 30
Most banks expect AI AI is reshaping the very foundations of financial
competition—from how trust is built, to how value is
and GenAI to lower costs. delivered, to who controls the customer relationship.

But the real disruption

runs deeper.

BOSTON CONSULTING GROUP FOR BANKS, THE AI RECKONING IS HERE 4

Traditional moats are being dismantled. Historically, Current profit models could also come under strain.
banks benefited from complexity. Customers stayed put, As AI-driven underwriting and real-time credit risk
pricing structures weren’t always clear, and financial assessment increase pricing transparency, they will reduce
products were tied to proprietary distribution channels. AI the margins banks can charge on loans.
is eroding these advantages in a number of ways:
Traditional advisory models will be disrupted as AI
• AI-powered agents will optimize financial decisions streamlines many aspects of portfolio management and
in real time, making it easier for customers to switch financial planning. Wealth management and commercial
providers and find better deals. Banks that once relied bankers will need to offer value that goes beyond what AI
on stickiness will need new ways to earn loyalty. can deliver and turn advisory into a differentiator.

• AI-driven transparency will expose rate structures, Fee-based transactional services will also face pressure.
fees, and lending terms in real time, eroding pricing AI-powered payment networks and embedded finance
power based on opacity. Banks will increasingly need players will pull more volume into ecosystems outside the
to compete on financial value—offering transparent traditional banking framework, forcing banks to rethink
pricing—as well as intangible value: the timeliness and their role in the value chain.
quality of their advice and how well they understand and
anticipate customer needs. Caution is costing time. Despite broad enthusiasm,
many banks remain wary of AI, particularly GenAI, and its
• AI-led financial decision making is shifting control current limitations, including its ethical and responsible
from banks to digital platforms that act as financial use. Large language models (LLMs) can “fantasize”—that
gatekeepers. GenAI is a hyper-accelerant in this is, generate non-factual output—when not tightly
evolution—enabling more autonomous, seamless, and controlled. They can also struggle with real-time data
personalized experiences that pull activity away from feeds, a critical gap in some financial applications. These
traditional banking channels. Agentic AI will amplify concerns have encouraged some banks to tread cautiously.
these changes, making it even harder for banks to own
the customer relationship. Investment levels reflect this hesitation. BCG’s AI Radar
found that one in three companies plans to spend over $25
million on AI in 2025, and some will spend in the range of
0.5% to 1% of revenues. But too much of this funding is
going toward isolated productivity improvements rather
than broader transformation. (See Exhibit 1.) Such
AI-led financial decision making investments suggest that many banks are still playing it
safe rather than positioning AI as a competitive
is shifting control from banks differentiator. In addition, while they are enthusiastic about

to digital platforms that act as AI’s potential, 60% of banks haven’t defined financial
performance indicators to track impact. Without clear
financial gatekeepers. metrics to ensure ongoing strategic alignment, they won’t
generate the ROI they need.

BOSTON CONSULTING GROUP FOR BANKS, THE AI RECKONING IS HERE 5

EXHIBIT 1

Most Players Are Not Yet Prioritizing Transformative Investments

Share of AI investments (%)

Invent
company-level innovations
27 core to the business

Deploy
initiatives focused on 44
individual productivity

Reshape
29 critical functions by improving
process-level productivity

Source: BCG analysis.

GenAI has moved past niche applications and is expanding

rapidly into core financial workflows, especially at the
Too much funding is going interface between institutions and customers. AI-powered
tools now support autonomous chat agents that go beyond
toward isolated productivity predefined scripts, real-time loan approvals, and
improvements rather than automated processing of submitted documentation.

broader transformation. The window to get ready for these changes is closing.
Within the next few years, and certainly by the end of the
decade, the banking landscape will look fundamentally
different. Leaders need to be modeling what this shift
means for their institutions—and defining the role they
intend to play in it.

BOSTON CONSULTING GROUP FOR BANKS, THE AI RECKONING IS HERE 6

Rethink, Retool, and
Reclaim Advantage
As AI reshapes banking, Meeting the moment will take more than pilots and point
solutions. It demands foundational changes—
leaders face a critical strengthening core pillars and, in some cases, rebuilding
them—to secure advantage for the decade ahead.
mandate: rethink strategy, Upgrade your strategy. Banks need to consider their AI
retool technology and data, vision and use it to chart their strategic positioning. Not all
will win in the same way. They need to identify where AI
upgrade governance, and can unlock durable advantage for themselves, then build
the infrastructure, talent, and partnerships to realize it.

prepare the organization. Equipped with those insights, banks can then consider
their strategic positioning and ask themselves, “Where can
we secure defendable advantage in the medium term?”
Three general models are emerging: the utility provider
focused on operational efficiency, the open-architecture
bank curating personalized financial products, and the
financial marketplace connecting customers with diverse
financial services. (See Exhibit 2.)

Each model leverages AI differently, whether it is to

optimize operations, recommend tailored products, or
create seamless, trust-based ecosystems. But all share a
common thread of moving beyond traditional lending.
Leading banks will create intelligent, customer-centric
platforms that generate value through data,
personalization, and strategic partnerships.

BOSTON CONSULTING GROUP FOR BANKS, THE AI RECKONING IS HERE 7

EXHIBIT 2

AI Will Play a Significant Role in Determining Where Banks Can Win

Core components of future business models AI vision (illustrative) HR strategy (illustrative)

Utility provider Heavily agentic Delivery efficiency

Focusing on scale and efficiency, provide core financial Go all-in on wide-ranging Higher volumes with fewer
services while third-party platforms handle customer automation, leveraging roles/FTE and cost savings
interactions. Profitability depends on volume, AI agents to monitor from productivity gain.
not direct customer ownership. and execute services.

Open-architecture bank Heavily insight-driven Sales effectiveness

Retain customer relationships while distributing third-party Invest in broad mix of Sales roles augmented by
financial products. Revenue shifts from net interest income predictive AI, GenAI, and (Gen)AI, with automation
to commission- and fee-based earnings. Winning requires customer-facing agents to of basic services.
using AI-driven customer insights to curate and recommend drive deep insight and
the right financial products. empower bankers and
customers.

Non-tied agent Balanced AI toolkit Platform organization

Evolve into financial marketplace, offering seamless access Balance investments in Flattening of the organization,
to providers (including non-banking). Business model relies platform automation with cross-functional teams,
on transaction fees and partnerships rather than lending investment into customer largely transversal setup.
margins. Success depends on trust, engagement, analytics and banker
and AI-powered curation and personalization. AI empowerment.

Source: BCG analysis.

Note: FTE = full-time equivalent.

Put AI at the center of tech and data. Making AI work • Data availability, not just accuracy, defines AI
at scale requires rethinking the architecture itself. This performance. Most AI failures in banking aren’t
demands changes across tech, data, and infrastructure: about the models—they’re about slow, incomplete,
or fragmented data. Unlocking AI’s full potential
• Workflow integration requires deep orchestration. requires addressing outdated systems and IT shortcuts
As banks evolve their AI capabilities, the challenge has (often referred to as technical debt), setting up strong
shifted from developing specialized models to integrating governance, and enabling efficient data integration
them intelligently. Orchestration matters, and GenAI across cloud and on-premise environments. LLMs
makes this nonnegotiable. (See Exhibit 3.) Banks will take a central role in banking AI, but they won’t
must design routing mechanisms that direct specific be sufficient. Many financial tasks are simply too
information to the best-fit model while also integrating specialized to rely on broad, general-purpose models,
proprietary data through techniques like retrieval- even when these are customized for particular domains.
augmented generation (RAG) and domain-specific small The use of purpose-built SLMs trained on specific data
language models (SLMs). Orchestration will become even for targeted GenAI applications will be key. Platforms
more critical as agentic AI use expands so that banks and orchestration systems that can optimize the use of
can coordinate decision execution as well as information LLMs and SLMs across AI-driven functions—bridging
flows. Responsible AI governance must also become current data silos—will also be needed.
integrated. Today, AI models are typically reviewed
individually by internal watchdogs. But as financial
institutions develop increasingly complex ecosystems—
with dozens or even hundreds of interconnected and
potentially autonomous models—banks will need holistic
oversight. This includes robust frameworks to stress-test
model interactions, identify emergent risks, and manage
intricate system interdependencies.

BOSTON CONSULTING GROUP FOR BANKS, THE AI RECKONING IS HERE 8

EXHIBIT 3

(Gen)AI Tech Architecture Requires New Capabilities

Smart Conversational apps (chat, search) Cognitive apps (e.g., Copilot)

Cognitive apps builder
Business (languages, libraries, etc.)
Layer Cognitive services (speech, language, text, image)

Transparency & Content moderation & Observability/ Ops and

Guardrails
ethical guidelines bias management feedback management Monitoring

Orchestration Prompt flow Agents Chains PromptOps

(Gen)AI Model garden

First-party foundation Third party Open-source Other small
LLMOps
models & pre-trained APIs models models models
Layer
Embeddings MLOps

Logging &
Model platform Feature store Model hosting/activation
monitoring Security Integration

Data products

Operational data services

Data Repository & storage (data lake, data warehouse, etc.) Prompt and output, etc. (data governance, data
Layer quality, DataOps, etc.)
Multimodal ingestion,
Distribution & integration (batch & stream processing, etc.)
retrieval, etc.

Core Transaction Layer

Specialized hardware
Infra and Cloud Layer Public cloud Private cloud
(GPU, TPU)

New architecture elements

Source: BCG analysis.

Note: LLMOps = large language model operations; MLOps = machine learning operations.

• Core layers must modernize. Most banking systems • Hybrid infrastructure is essential. Today, AI systems
are a technological patchwork that obstructs the dynamic, can flag risks, surface insights, and suggest pricing
real-time, and unstructured capabilities essential for changes—but most don’t trigger real-time adjustments.
innovative AI applications. Simply adding AI components This must change. There are many opportunities where
to existing infrastructure won’t work. Leading institutions predictive and agentic AI can work together to propose
are demonstrating a new approach. Commonwealth Bank an action and then implement it without exposing the
of Australia, for example, has implemented an event- bank to risk. Fully personalized marketing interactions
driven architecture and an AI-powered transaction core. are one example. For these opportunities to expand,
These allow for real-time fraud detection and response, infrastructure needs to be hybrid. It must cut across
contributing to a 50% drop in scam losses and a 30% on-premise, cloud, and edge environments to enable
decrease in customer-reported fraud. high degrees of modularity and the widespread use of
application programming interfaces and micro-services.

BOSTON CONSULTING GROUP FOR BANKS, THE AI RECKONING IS HERE 9

Own the AI governance agenda. With regulation still Banks also need to be thinking one step ahead and
catching up to AI’s rapid evolution, banks are being asked integrate AI governance into their business strategy. Most
to govern in real time, without a clear playbook. The EU AI need to develop risk management frameworks that go
Act is setting the most comprehensive standards to date, beyond checklists so they can better address explainability,
applying across the entire AI value chain. Banks must now accountability, and bias detection—as well as prepare for
ensure oversight not only for their own models but also for an uptick in AI-driven financial crime. Standard Chartered,
any third-party systems they rely on. While the act isn’t for instance, is already investing in an AI platform that can
finance-specific, its breadth makes AI governance a identify compliance failings and potential fraudulent
business-wide responsibility, not just a compliance task. behavior. This initiative is part of the bank’s efforts to
develop responsible AI risk management frameworks,
In the US and UK, regulators are incorporating AI oversight particularly for bias detection in lending and credit
into existing financial rules rather than building separate decisions, in response to evolving regulations.
frameworks. Still, varying state laws in the US are likely to
cause significant fragmentation in how those policies are Ready the organization. Today, a typical bank might have
applied. US agencies, including the Federal Reserve, are 15% of staff in the front office, 10% in control, 10% in
focused on AI model risk, bias detection, and explainability, operations, 20% in corporate, 20% in the middle office, and
particularly in lending and credit. The UK’s Financial 25% in technology. How might this composition change with
Conduct Authority and Prudential Regulation Authority are GenAI and agentic AI? The answer will lie in accordance with
integrating AI risk frameworks into broader financial each bank’s AI vision and business strategy. Those focused on
governance. These approaches give banks more flexibility utility services may maintain a larger operations workforce,
but also shift greater responsibility onto them to ensure while others may prioritize advisory and technology.
that AI models are reliable, fair, and auditable.
Few employees are being prepared for these shifts. BCG
Other changes are also brewing. The Bank of England, for research found that two-thirds of financial institutions have
example, is exploring how to incorporate AI into its stress- difficulties hiring AI talent, and fewer than a third have
testing regime—assessing whether trading models and upskilled even a quarter of their workforce. The issue isn’t
risk algorithms could amplify financial instability. If just hiring AI specialists—it’s ensuring that decision
adopted, banks could eventually be asked to demonstrate makers and oversight teams have the skills to assess,
both balance sheet resilience and how AI behaves under challenge, and apply AI outputs effectively. Many frontline
extreme conditions. This secondary requirement would and control teams still struggle to interpret AI-driven
necessitate completely new technical capabilities. decisions, let alone justify them to regulators.

Early engagement with regulators will be key to managing Some banks are exploring solutions. JPMorgan Chase has
the direction of this journey. For now, however, uncertainty implemented a GenAI tool called LLM Suite, accessible to
remains the biggest drag on AI adoption in finance. In our 200,000 employees, including CEO Jamie Dimon. The bank
research, 61% of institutions cite regulation as a top offers training programs and leverages superusers to assist
concern. But waiting for clarity isn’t a strategy. Indeed, colleagues in integrating AI tools into their workflows.
most supervisory authorities are looking for banks to BBVA has partnered with the University of Navarra to
engage with them and uncover areas where effective AI launch a training initiative targeting over 150 top
regulation can remove barriers to growth. managers. This program focuses on the use of GenAI to
improve executives’ productivity by optimizing their
strategic decision making and daily operations.

Supervisory authorities are looking for banks to engage

with them and uncover areas where effective AI regulation
can remove barriers to growth.

BOSTON CONSULTING GROUP FOR BANKS, THE AI RECKONING IS HERE 10

Act Now
The choices banks make Get started with no-regrets moves. Leaders should
prioritize depth over breadth and move fast where the upside

now will define the next is clear. Here are three steps they should take this year:

generation of leaders.
• Systematically evaluate where AI can drive ROI.
High-value use cases exist across every bank function
from operations to wealth management. Assessments
Success will depend on should drill into the numbers, denoting expected value
and cost as well as the underlying assumptions. Include
a disciplined approach: qualitative benefits as well, such as faster processes,
better insights, and a better user experience.
focusing AI on measurable • Connect use cases to processes. Use case selection

returns, embedding it should inform a wider workflow analysis to drive process

redesign and integration across the relevant function or

into decision making, and organization. This mindset is the key to scaling—and
allows organizations to see impact faster and build it

adapting quickly as new

more sustainably.

• Set clear outcomes and accountabilities.

opportunities emerge. Momentum is everything. Clear goals aligned to
business outcomes sustain momentum. Aim for quick
wins and decisive impact within a three-year horizon to
unlock faster ROI.

BOSTON CONSULTING GROUP FOR BANKS, THE AI RECKONING IS HERE 11

Drive alignment through visibility and intent. AI will Lead from the top. In many banks, AI is still viewed as
reshape core workflows, decision rights, and value streams the exclusive domain of the CTO or innovation lab. That
across the bank. Change at this scale demands tight mindset is now the primary blocker to scale. This is how
alignment. Here’s how to achieve it: leaders can drive the organizational, financial, and talent
decisions that AI maturity demands:
• Track what matters. Many banks lack meaningful AI
performance indicators. Shift from counting pilots to • Leverage the full power of the CEO. The CEO should
measuring adoption, time-to-impact, and risk-adjusted define the AI strategy, set KPIs, and ensure that execution
ROI. Build enterprise dashboards that surface business- is disciplined. Banks with many independent divisions
relevant insights—not just model technical metrics. or front-to-back opportunities may opt for a divisional or
centrally coordinated approach, but without clear top-
• Clarify what’s working and where to go next. AI- down leadership, AI adoption will remain fragmented.
driven change is moving faster than most banks are
planning for and in ways that are not predictable. CEOs • Get serious about data governance. Banks know
must condition their organization to keep their eye on that data governance is critical. Few are actually fixing
emerging developments and model the implications— it. CEOs must push past surface-level coordination and
and their response to them. In our experience advising mandate the standards, ownership, and funding required
institutions, we’ve seen that progress often stalls as a to make trusted data usable at scale.
result of limited visibility. CEOs must make sure that
senior leaders across the organization are brought into • Own the talent question. Competing in an AI-first world
the overall AI transformation. And they should employ requires internal fluency, not just external partnerships.
formal and informal channels to prioritize ongoing CEOs must back the development of in-house expertise
communication so that managers from the top levels on across product, tech, and operations and create the space
down understand what’s working, what’s ready to scale, for new roles and ways of working to take hold.
and what to fund.
AI is already reshaping banking. The real challenge now is
• Move capital like you mean it. Shifting from pilots to ensuring that institutions harness its potential thoughtfully
transformation requires reallocation of real budget and and strategically.
talent. Building the tech and foundations to scale eats
up to 60% of AI investment in the first two to three years. The age of incrementalism is over. In banking, as in every
But the longer banks wait, the larger this gap gets—to industry AI touches, the institutions that thrive will be
the point where it cannot be closed. Banks must end those that rethink not just tools and workflows—but value,
low-potential pilots, double down on proven use cases, control, and differentiation from the ground up. That
and ensure that foundational AI capabilities are funded means anchoring AI strategy in business strategy, investing
as core infrastructure. in the technical and human foundations to support it, and
embedding clear lines of accountability. Leadership
matters. So does urgency.

The opportunity is enormous—and fleeting. Banks that

move decisively will define the future of financial services.
The rest will compete for what’s left.

BOSTON CONSULTING GROUP FOR BANKS, THE AI RECKONING IS HERE 12

About the Authors

Saurabh Tripathi is a managing director and senior Stiene Riemer is a managing director and partner in
partner in Boston Consulting Group’s Mumbai office and BCG's Munich office and the global lead for AI
global leader of the Financial Institutions practice. He can developments in financial institutions. You can reach her
be reached at [email protected]. at [email protected].

Matteo Coppola is a managing director and senior Kirsten Rulf is an associate director and partner in the
partner in the firm’s Milan office and global practice leader firm’s Berlin office and a core member of the Technology
for Risk and Compliance. You can reach him at and Digital Advantage practice. You can reach her at
[email protected]. [email protected].

Jürgen Rogg is a managing director and senior partner in Christian Schmid is a managing director and senior
BCG’s Zurich office and global leader of the Digital and partner in BCG's Zurich office and a core member of BCG's
Technology in Financial Institutions topic. You can reach Public Sector, Financial Institutions, and Technology
him at [email protected]. Advantage practices. You may contact him by email at
[email protected].

Michael Strauß is a managing director and senior partner For Further Contact
in the firm’s Cologne office and leads BCG's Financial
Institutions practice in Central Europe and GenAI in Europe. If you would like to discuss this report, please contact
You can reach him at [email protected]. the authors.

BOSTON CONSULTING GROUP FOR BANKS, THE AI RECKONING IS HERE 13

Boston Consulting Group partners with leaders in business
and society to tackle their most important challenges and
capture their greatest opportunities. BCG was the pioneer
in business strategy when it was founded in 1963. Today,
we work closely with clients to embrace a transformational
approach aimed at benefiting all stakeholders—empowering
organizations to grow, build sustainable competitive
advantage, and drive positive societal impact.

Our diverse, global teams bring deep industry and functional

expertise and a range of perspectives that question the
status quo and spark change. BCG delivers solutions
through leading-edge management consulting, technology
and design, and corporate and digital ventures. We work
in a uniquely collaborative model across the firm and
throughout all levels of the client organization, fueled by
the goal of helping our clients thrive and enabling them to
make the world a better place.

For information or permission to reprint, please contact

BCG at [email protected]. To find the latest BCG
content and register to receive e-alerts on this topic or
others, please visit bcg.com. Follow Boston Consulting
Group on LinkedIn, Facebook, and X (formerly Twitter).

© Boston Consulting Group 2025. All rights reserved. 5/25