Artificial intelligence is evolving at a remarkable pace, driving a wave of investment across industries. According to the World Economic Forum, financial services alone are projected to spend close to $100 billion on AI by 2027. Yet the real issue isn’t the size of the investment, but the outcomes it produces.
As Microsoft CEO Satya Nadella has noted, success shouldn’t be defined by hitting AI milestones, but by whether it accelerates productivity and economic growth. The ultimate measure is straightforward: does AI deliver tangible benefits for organisations – and broader prosperity for society?
The challenge: efficiency vs growth
AI adoption in banking is now widespread, though maturity varies. Much of what is live today is about doing more with less: automating repetitive work, reducing error rates and compressing cycle times as demonstrated by recent Bank of England and FCA research.
Payment processor Stripe is a clear example. By moving from hand-engineered rules to a self-supervised foundation model that analyses billions of payment sequences, the company lifted fraud detection rates on card-testing attempts from 59% to 97%, directly protecting revenues and improving margins.
Meanwhile, generative systems can change service economics at scale. Brazilian Nubank’s OpenAI-powered co-pilots now handle more than 2 million customer chats a month, freeing agents for higher-value work and improving unit economics in a low-fee retail market.
These cases prove AI works when applied to structured, data-rich workflows with clear objectives. Yet they remain focused on efficiency and compliance rather than growth – the greater opportunity lies in using AI to drive new revenues, deepen client relationships and create markets.
For example, AI is now a sales tool, not just a back-office assistant. JPMorgan’s Coach AI supports the bank’s advisers by anticipating client questions and surfacing relevant research, contributing to a 20 per cent rise in gross sales in asset and wealth management between 2023 and 2024. Initiatives like Google’s AP2 protocol point to a near future where agents not only reason and act but also transact. And by embedding AI nudges into its NAV Planner and Everyday app, DBS has changed how customers manage their money. Users who receive these prompts save roughly twice as much and invest around 5 times more than those without.
For DBS, that behaviour shift translates into tangible growth. The bank estimates its portfolio of AI applications has generated S$750mn ($580mn) in 2024 and is on track to exceed S$1bn in 2025 – delivering value that is flowing into deposits, investment fees and insurance distribution.
Google, Mastercard, Visa and others are laying the emergent foundations for agent-to-agent payments, such as Google’s AP2 protocol, to enable agents to transact securely. Though early, these initiatives point to a near future where agents not only reason and act but also transact – enabling new opportunities in IoT settlements, supply chains and machine-to-machine commerce.
Building on this are exciting new growth opportunities where agentic AI systems interact safely and effectively with other technology-enabled financial innovations, such as open banking, open finance and tokenisation.
These cases show that growth emerges when AI is tied to revenue-bearing workflows – advisory, cross-selling, embedded finance – and when institutions invest in trust, data quality, data portability and governance to scale beyond pilots.
Growth emerges when AI is tied to revenue-bearing workflows – advisory, cross-selling, embedded finance – and when institutions invest in trust, data quality, data portability and governance to scale beyond pilots.
Growth with AI remains out of reach for most banks
For most banks, the leap from efficiency to growth remains elusive. Familiar barriers persist: fragmented data locked in silos, legacy IT systems, weak feedback loops and objectives too vague to sustain investment. Faced with these constraints, institutions default to using AI for incremental savings rather than bold innovation.
Recent MIT research found that only 5% of generative AI pilots scaled. Even the largest tech firms are still wrestling to show payback on big AI bets(The Pragmatic Engineer, 2025). The Financial Times’ study of S&P 490 filings on the impact of AI points to a similar challenge: defining the benefits of AI while countering new risks.
1
Accountability
The first risk is accountability. When AI systems misfire – a biased credit decision, a faulty recommendation or a trading error, who bears responsibility? The model provider, the AI vendor that wrapped it or the deploying institution?
2
LLMs and agentic AI
Second, large language models (LLMs) and agentic AI add new risks into the mix: from inconsistent, unreproducible outputs to explainability gaps that worry regulators and consumers alike. The largest, most complex systems founded on billions of parameters are the least understandable by humans.
3
Cyber attacks
Third, greater autonomy makes these systems prime targets for novel cyber attacks from data poisoning and prompt injection to deepfakes and even designing whole new attack strategies using AI, which are outlined in the Mitre Atlas taxonomy.
Without cleaner data, stronger infrastructure, clear accountability and explainable models, most financial institutions will continue to find the gap between AI’s promise and growth too wide to bridge.
From pilots to scale – enabling growth through AI
Banks cannot cross this gap alone. The ecosystem of AI vendors spanning compute, data management, analytics, model development and workflow automation is becoming as critical as the AI systems themselves.
Cloud providers deliver the compute backbone that enables banks to train and run models, often on Nvidia-powered infrastructure. Data management, analytics and governance vendors enable audit trails and data lineage.
Meanwhile, firms such as HuggingFace provide open-source libraries, pretrained models and frameworks that enable financial institutions to adapt to their domain needs. Then there are workflow automation platforms that help to embed intelligence into everyday processes, from onboarding to risk reviews.
Together these firms represent a broader, fast-evolving ecosystem of AI enablers that provide the components to move banks from pilots to scale.
The ecosystem of AI vendors spanning compute, data management, analytics, model development and workflow automation is becoming as critical as the AI systems themselves.
Closing the evidence gap for AI in financial services
What the industry still lacks is reliable evidence. Despite the hype, few benchmarks exist on where AI is being deployed, what returns are being generating, and how AI is governed. This leaves institutions struggling to separate substance from promise.
Due to be published in Q1 2026, the AI in Financial Services 2030 Global Surveys – led by the Cambridge Centre for Alternative Finance in collaboration with the IMF, BIS, World Economic Forum, World Bank, IDB, CGAP and Financial Innovation for Impact – aims to close that gap.
It will deliver the first integrated global dataset on adoption, impact and governance, giving financial institutions the evidence they need to move beyond pilots and scale with confidence. There are three global surveys to explore how AI is being adopted and impacting:
- financial authorities
- industry
- AI vendors
Only by moving beyond efficiency plays to new measurable value creation – through new revenues, deeper client relationships and market-shaping innovation – will AI in finance realise its full transformative potential.
Only by moving beyond efficiency plays to new measurable value creation – through new revenues, deeper client relationships and market-shaping innovation – will AI in finance realise its full transformative potential.
This article originally appeared in The Banker on 7 October 2025, titled ‘Why banks struggle to turn AI efficiency into growth’.
Featured authors
Kieran Garvey
AI Research Lead for the Cambridge Centre for Alternative Finance
Bryan Zheng Zhang
Co-Founder and Executive Director, Cambridge Centre for Alternative Finance; Executive Chair of Financial Innovation for Impact
Read more about the AI in Financial Services 2030 Global Surveys
Related content
An estimate by the World Economic Forum suggests financial services will spend nearly $100bn on AI by 2027.
Microsoft’s Satya Nadella says progress in AI should be judged by impact on productivity and growth, not milestones.
Bank of England and FCA research suggests much of what is being done is about doing more with less (for example automating repetitive work).
Stripe lifted fraud detection rates on card-testing attempts from 59% to 97%, directly protecting revenues and improving margins
Nubank’s AI co-pilots manage over 2million chats monthly, freeing agents for higher-value work and improving economics.
JPMorgan’s Coach AI helps advisers anticipate client needs and access research, driving a 20% sales increase in asset and wealth management.
Google’s AP2 suggests a future where AI agents can not only think and act but also complete transactions.
DBS uses AI nudges to boost customer saving and investing, driving significant growth and revenue.
MIT research found only 5% of generative AI pilots scaled.
FT’s study of S&P 490 filings shows the challenge of defining AI benefits while managing risks.
Mitre Atlas outlines how attackers exploit AI systems, detailing tactics like data poisoning, adversarial inputs, and model manipulation.
Related articles
AI and technology
Innovation to impact: technology, governance and regulation
As governments worldwide grapple with the accelerating pace of technological change, the central challenge is no longer technical but institutional: how to translate digital innovation into citizen-first outcomes at scale. The obstacles lie in execution – requiring effective governance, adaptive regulation, and institutional transformation. In this piece, Carlos Montes (Lead – Cambridge Innovation Hub for Prosperity, Montek Singh Ahluwalia (citizen-first public servant and economic reformer), and Pavle Avramovic (Head of Market and Infrastructure Observatory at the CCAF), examine how public institutions can move from policy ambition to real-world impact by building innovation-ready systems that prioritise citizens over processes
AI and technology
From innovation delta to regulatory singularity: how innovative regulatory systems can help regulation keep pace with financial innovation
The global financial system is undergoing a profound transformation driven by digital innovation. Yet, the pace of regulatory innovation and the scale of digital transformation within public authorities remain uneven and comparatively insufficient. This article, by Bryan Zhang, Co-Founder and Executive Director of the Cambridge Centre for Alternative Finance, examines the widening innovation gap between innovators and regulators, and outlines potential strategies to foster and accelerate system-level innovation within central banks and financial regulators across both developed and developing economies.
Finance and accounting
Transition finance: bridging the climate investment gap
Philippa Martinelli, Head of Climate & Transition Finance at the Cambridge Centre for Alternative Finance, explores the vital role of finance in tackling climate change. The article highlights the scale of the challenge, the urgency of transition finance, and the investment shifts needed to build a low-carbon, resilient global economy.