In 2026, financial services institutions (FSIs) are moving far beyond AI experimentation to embrace widespread AI deployment. As organisations target AI-driven productivity gains and agentic systems increasingly automate the back office, it’s clear that leaders from across the sector are at a critical crossroads. Ahead of FStech’s The Future of AI in Financial Services – taking place at London Hilton Tower Bridge on 6 October – associate editor Rory Bathgate speaks with some of the event’s speakers to unpack their thoughts and observations on the sector’s future.

AI adoption in financial services has accelerated rapidly over the past few years. From the early days of experimentation and cautious adoption, through to the explosion of generative AI into the market in 2022 and today’s widespread AI tool rollout, FSIs have been forced to navigate a shifting landscape as full of opportunity as it is fraught with risk.

The extent to which the sector has embraced AI systems, even in the face of regulatory requirements and risk appetites that hold it back from the fastest of adoption curves, is noteworthy.

An April report by the Cambridge Centre for Alternative Finance (CCAF) found that 81 per cent of financial services firms are already adopting AI to some extent, with 71 per cent having adopted generative AI.

Agentic AI is changing the speed and scale of investment, with many banks now turning to autonomous systems for back-office tasks such as credit memo preparation and increasing interest in agentic commerce at the global level. The CCAF report found that 52 per cent of industry respondents are adopting agentic AI systems.

This also brings its own challenges, with institutions such as the Bank of England having warned that existing regulation may not be able to encompass the systemic risks of AI agents.

From experimentation to embedded AI

Vibhor Narang, lead for payments and treasury solutions, UK & Europe, at Standard Chartered, tells FStech that the sector is finally shifting from AI experimentation to measurable AI outcomes.

“Immediate priorities include intelligent bots, document automation, regulatory reporting, cash-flow forecasting, analytics and reconciliation,” he says. “At Standard Chartered, we are focusing on practical applications across corporate banking, including trade-document checks, client research and liquidity-solution design. The real prize is not automation alone, but better, faster and more consistent financial decision-making.”

This is a focus shared by Tom Martin, business platform lead for Economic Crime Prevention at Lloyds Banking Group.

“We are now looking at where it can deliver clear, measurable value across the business, whether that is improving the customer experience, helping colleagues work more effectively, tackling fraud or simplifying complex operations,” he tells FStech.

“It is becoming an important part of our wider transformation, with applications across customer service, software engineering, fraud prevention and financial guidance.”

Hellen Beveridge, head of AI Governance & Ethics at AXA UK, tells FStech that in practical terms this means FSIs are looking to boost staff productivity and harness the technology to fight cyber threats and fraud.

“As financial criminals become more advanced, institutions are recognising that traditional methods are insufficient, prompting a surge in AI-driven fraud detection systems that offer real-time, adaptive and highly accurate identification of suspicious activities,” Beveridge says.

“Companies are also looking to make better use of their data. Financial services firms have huge amounts of information, and AI is creating new opportunities to turn it into insights, improve decision-making and deliver better customer experiences.”

Fraud detection and data analytics are two areas where FSIs can reap clear value from AI, as the productivity gains of large language models (LLMs) overlap with the tried-and-tested benefits of ‘traditional’ AI approaches such as machine learning.

But as mentioned above, FSIs are also increasingly interested in the rollout of more autonomous systems such as AI agents capable of completing multi-step workloads. Martin tells FStech that while these systems are a step ahead for productivity, they must be grounded in the same foundations as any other piece of technology.

“None of that works without the right foundations,” he says. “Investment in data, platforms and skills matters just as much as the technology itself. To use AI effectively and safely, firms need good-quality data, clear governance, the right operating model and colleagues who understand how to work with it.”

As AI becomes a core product offering of many cloud and software providers, it is also becoming increasingly embedded into software by default. Beveridge notes that for some FSIs, this leads to a kind of passive AI adoption where the technology becomes a part of a service they already use.

To an increasing degree, she adds, both active and passive adoptees are experiencing a “growing sense of realism” around what AI adoption means for dependency on tech firms.

“The conversation is starting to move beyond ‘what can this do?’towards ‘what happens if this becomes critical to how we operate?’” Beveridge explains. “Organisations are becoming much more interested in resilience, portability and avoiding dependence on technologies whose future economics may look very different from today's.”

Control as a marker for success

Much of the conversation around AI in financial services to date has been around building strategies for AI adoption, the potential first-mover advantages available to FSIs with a higher risk appetite and the specific areas AI can add value.

Martin says this conversation is now shifting. “Over the next two years, the question will increasingly be not where AI can be used, but how it can be scaled safely and effectively,” he says.

“Generative AI has already shown what it can do. The next step is to build it more deeply into customer journeys and day-to-day operations, with clear controls over where and how it supports decisions.”

At Lloyds, he adds, this could look like AI agents deployed in tandem with human colleagues.

“They could help with routine tasks, bring relevant information to the surface and support customer interactions. In a regulated business, however, clear accountability, transparency and human oversight remain essential.”

Beveridge also acknowledges this need, noting that over the next few years she expects to see FSIs more rigorously evaluate AI deployment.

“The next phase is about embedding it into business processes in a way that's sustainable, governed and measurable,” she explains.

“There will also be a much greater focus on what happens after deployment. Launching an AI capability is one thing. Understanding how it's performing six months later is something else entirely. Monitoring, assurance and ongoing oversight will become far more important.”

Narang notes that the need for control and reliable outputs will put a natural limiter on the extent to which AI agents are rolled out in the enterprise. Although AI will become even more integrated into core workflows, he tells FStech, no financial organisation is likely to run them without constant oversight.

“Intelligent agents may recommend liquidity or funding actions within predefined policies, but human judgement will remain central,” he says. “The next phase of AI will be measured not by novelty, but by how quietly it improves the speed, quality and consistency of financial decision-making.”

Beveridge sums this up by noting that AI adoption in financial services can only be scaled with the right governance in place.

“The organisations that get the most value from AI won't necessarily be the ones deploying the most models,” she says. “They'll be the ones that can scale adoption while still maintaining control.”

Hurdles to overcome

The understanding that bad data inevitably results in bad AI outputs, often summarised as ‘garbage in, garbage out’ has become something of a mainstay for enterprises adopting hallucination-prone generative AI models. But it goes without saying that in a sector as highly regulated as financial services, there is no room for numerical errors within AI outputs.

Martin acknowledges this, drawing on the example of Lloyds’s own AI adoption. “Customers rightly expect Lloyds to use AI accurately, and trust depends in part on the quality of the data behind AI,” he says. “An AI system can only be as reliable as the information it uses, so sound data management, security and risk controls are fundamental.”

Beyond the accuracy of AI outputs, FSIs also face an uphill battle with legacy systems and data hygiene. The CCAF report found that data availability and quality remain a major barrier to AI adoption, cited by 40 per cent of financial services sector respondents.

Narang echoes these concerns in his assessment of the situation many FSIs face. “The principal hurdles are fragmented data, legacy integration, explainability, cyber resilience and accountability,” he explains.

In addition to these technological barriers, FSIs face sector-specific challenges that do not align with the ‘move fast and break things’ approach promoted by big tech.

“Treasury decisions directly affect liquidity, payments and financial risk, so organisations need clear approval thresholds, audit trails and human intervention,” explains Narang. “Third-party concentration is another emerging concern. Trust will be the licence to scale: innovation must be explainable, resilient and governed across the full AI lifecycle.”

Beveridge expands on third-party risks, noting that many FSIs are now forced to adapt to AI and create rules for adoption as suppliers begin to use the technology.

“Firms need to understand new technology risks, assess how data is being used, review transparency and control mechanisms, and often update contractual arrangements to reflect AI-specific considerations,” she says. “In some cases, the governance effort associated with a supplier's AI capability can be as significant as that of an internally developed solution.”

For Martin, adaptation is rooted in people and skills. He tells FStech that organisations must strive to bring their people along with them, if they are to deploy AI effectively.

“Leaders, technology teams and operational colleagues all need to understand how to use AI well and where human judgement still matters,” he says. “At Lloyds, continuing to build that understanding across the organisation is an important part of adopting the technology responsibly.”

Regulatory uncertainty and customer trust

Regulators have been clear in their demands for greater powers to control deployment of the technology in financial services. In June, the Financial Stability Board (FSB) warned that AI models present unique risks for FSIs and called for more stringent regulation of the sector.

In July, the UK’s Financial Conduct Authority (FCA) said it is looking into leveraging stronger regulatory powers over the use of AI in financial services, with its executive director Sheldon Mills stating that regulators are in an “arms race” to keep up with potential risks.

Narang says that he expects the regulatory landscape for AI in financial services to markedly change over the next 12 to 24 months. He adds approaches will differ however, pointing to the broad spectrum of regulation already in place including the EU AI Act, US federal and state frameworks, Singapore’s Model AI Governance Framework and the UAE Central Bank’s Responsible AI Guidance.

“AI may be global, but its licence to operate will remain local,” he said. “International banks must combine consistent enterprise governance with flexibility to meet jurisdiction-specific requirements across their networks.”

Beveridge says that regulators are focused on the same things as FSIs, such as accountability, data protection, transparency and resilience but more clarity is still needed. This will come in the next few years, she tells FStech, along with greater scrutiny of AI adoption.

“Regulators will increasingly want evidence that firms understand where AI is being used, what risks it creates and how those risks are being controlled,” she says. “I also expect greater attention to the responsibilities of technology providers, particularly as organisations find themselves adopting AI through third-party products and services rather than purely through their own development programmes.” 

“Ultimately, the organisations that are investing in strong governance now are likely to be in a much better position regardless of exactly how the regulatory landscape develops.”

Martin tells FStech that he expects regulation to continually evolve and hopes to see FSIs given room to innovate, provided they are following a safe route to adoption.

“The next 12 to 24 months should bring clearer expectations on governance, transparency, accountability and risk management, particularly as more advanced and agentic systems begin to be used in practice. Our work on responsible AI puts Lloyds in a good position to respond as those expectations develop,” he says.

“Regulation does not need to hold innovation back. Done well, responsible AI frameworks give firms the confidence to adopt the technology safely. What matters is practical, proportionate oversight that supports innovation without compromising customer trust, transparency or accountability.”

There is no doubt that AI is being adopted at pace by the financial services industry. The change this could bring is significant, with manual tasks automated to a more meaningful extent, workers able to access better data as they need it, and FSIs empowered to protect their customers from fraud and cyber threats.

At the same time, FSIs will have to stick to the fundamentals including accountability, data protection and cyber resilience to ensure they stay on the right side of legislation, even as it evolves to meet the needs of future AI systems. A strong focus on governance and risk management will also keep options open for FSIs, by helping them to clearly assess the costs and benefits of AI systems and whether they come at the cost of third-party dependency.

If current trends persist, AI will become an even more critical part of the financial services sector in the years to come. Those that engage in the conversation now, sharing best practices with industry peers and cross-sector experts, stand to benefit the most.

Vibhor Narang, Tom Martin and Hellen Beveridge will be among the senior delegates at FStech's The Future of AI in Financial Services, taking place at London Hilton Tower Bridge on 6 October. The conference brings together an excellent roster of speakers from across the sector, including Standard Chartered, Lloyds Banking Group and AXA UK, to share how they are moving from AI experimentation to measurable, well-governed outcomes. Attendance is free for professionals working within financial institutions. Click here to learn more and to sign up.

---