Banks in rapid take-up of AI despite growing data security risks

Banks are increasingly leveraging artificial intelligence systems and tools as they look to offer customers a more personalised experience, streamline their operations, tackle rising levels of fraud and, ultimately, stay ahead of the curve in a highly competitive marketplace.

But without proper human oversight and guidelines, experts warn that AI — computer programs trained to replicate human cognitive functions such as learning, problem solving, decision making and creative thinking — could put customer data at risk, provide false or inaccurate financial information and even hamper banks’ net zero targets.

Disrupting all of banking

According to statistics from the Bank of England, 75% of UK-based financial firms have already implemented AI. Meanwhile, in the next three years, a further 10 per cent will invest in the technology, meaning it will have impacted virtually all of the sector.

One of these is Starling Bank, whose chief information officer, Harriet Rees, says the challenger bank is using the technology “across all departments”. In its customer service division, Starling has equipped its agents with AI tools that allow them to create call summaries and find relevant information quickly.

Consequently, she claims, customers now only need to wait 14 seconds to be seen by a customer service representative — a 60 per cent reduction compared to previous waiting times. And Starling’s customer service agents are saving themselves 8,000 hours of work each month, with the number of calls handed over to other specialists halved.

Elsewhere, Starling’s cyber security department uses automated fraud detection systems that have prevented authorised push payment fraud — scams that involve cyber criminals tricking victims into sending money to their bank accounts — by 16%.

In addition to adopting AI internally, Starling is also integrating the technology into its customer experience. For example, its Spending Intelligence tool uses AI to help customers better understand their finances and make better financial decisions.

The tool questions customers on how they spend their money via text and voice-based inputs, explains Rees. She adds: “Customers can find out how much they spent on takeaways last year, how much they spent on taxis last month or analyse their weekly grocery spending.”

ClearBank, the first new British clearing bank in 250 years, is also heavily reliant on AI. With the technology, the bank has automated the process of returning erroneous payments, analysing complex information, identifying and fixing common bottlenecks in its customer experience, and reviewing customer claims.

Ben Altieri, head of product within the client services and revenue solutions division at ClearBank, says these AI use cases have enabled the bank to “reduce friction, unlock capacity, and deliver faster, smarter outcomes”.

Traditional banks, too, see the benefits of AI usage. Lloyds’ head of AI, Rohit Dhawan, says they have the potential to “fundamentally reimagine” how banks operate. Like Starling, Lloyds has rolled out the technology across its customer service department — which now has access to a proprietary AI assistant, called Athena, for information gathering. Dhawan says the tool is “dramatically reducing search times and improving accuracy”.

As well as building its own AI tools, Dhawan and his team at Lloyds are also leveraging third-party tools like Google Cloud’s Vertex AI. The latter provides a single interface from which organisations can easily develop, train and roll out AI applications. According to Dhawan, it is helping Lloyds “scale innovation securely and responsibly”.

He tells The Banker: “These technologies are helping us reimagine customer engagement, streamline operations, and empower our colleagues with intelligent tools that make their work more impactful.”

Getting the most out of AI

For banks new to using AI, starting “small” is key. That’s the advice of ClearBank’s Altieri, who says that doing the opposite could result in AI projects becoming “stuck in pilot mode” and “held back by risk and regulation”.

Altieri explains that when ClearBank first began experimenting with AI systems, it used them for “high-value” but equally “low-risk” applications internally. By doing so, Altieri and his colleagues were able to demonstrate that the tech could deliver “measurable results” for the wider business, which resulted in support from the bank’s executive team and board of directors.

This sentiment is echoed by Gilles Chemla, a professor of finance at Imperial College Business School and co-director of its Centre for Financial Technology. He encourages gradual AI adoption, whereby banks deploy the technology for “specific use cases” that align with clear objectives and are carefully reviewed by human banking specialists.

To ensure AI projects are a success, Dhawan of Lloyds says banks must ensure they have the “right foundations” from the outset. First, he says adequate guardrails are needed so that AI applications are used safely, ethically and in line with banks’ compliance obligations.

Second, he recommends that banks adopt open-source IT systems — whose code is widely available for improved collaboration and interoperability — so they can leverage any AI tool. Finally, he urges banks to implement AI platforms that cater to everyone’s needs, rather than just tech experts. This is the key to ensuring widespread organisational uptake.

Because AI systems are trained on large volumes of data, taking steps to keep underlying datasets error free and well structured is also important. According to Rees of Starling Bank, this will help banks ensure their AI tools offer the “most effective” insights.

Once banks have a good grasp of AI and the correct systems in place, Rees says they can experiment with “more advanced or customer-facing applications”. But whether an AI application is simple or more complex, she says banks should always have a clear rationale behind it. She explains: “Any good technologist will tell you that you should start with the problem you want to solve, rather than looking to implement the technology for the sake of it.”

Overcoming AI’s pitfalls

Although AI is proving to be a powerful banking technology, it is not without downsides. One of the most significant is the data privacy and cyber security implications it poses.

Chemla of Imperial College London warns that “digital signatures, biometrics, and voice/image recognition can all be subject to AI-induced fraud”. With this in mind, he says human oversight is needed to ensure AI systems are secure and used safely.

However, this is where leveraging existing industry security frameworks and guidance like the Bank of England’s AI Baseline Guidance Review can help, according to Rees of Starling Bank. As a result, she says banks can push ahead with AI projects “responsibly and with confidence”.

Large language models like ChatGPT and Google Gemini are also susceptible to AI hallucinations, which is when they provide incorrect, fabricated or exaggerated responses.

To tackle this issue, Chinmoy Banerjee — president at IT consulting firm Hexaware Technologies — says banks should “establish strong validation and monitoring systems” that can “ensure the reliability and accuracy of their AI-generated outputs”.

As AI technologies consume considerable amounts of power, they could also detrimentally impact banks’ net zero targets. Therefore, Michael Conway, senior partner of UK banking at American tech giant IBM, says banks should use the technology in a sustainable manner.

He advises: “Instead of defaulting to big, proprietary, carbon-heavy and expensive models, it’s more cost and energy-efficient to really consider what you need from your model — and only include the necessary dataset and parameters.”

Banking AI projects can also be affected by staff pushback, which often stems from the belief that the technology will make their roles redundant. To avoid this, Dhawan at Lloyds advises banks to educate employees on how AI works and how it can benefit them in their roles.

Looking ahead at banking AI

Challenges aside, AI seems to have a bright future in the banking sector. In particular, Dhawan of Lloyds believes that agentic AI systems — which can make decisions without human input — could “radically transform productivity and customer experience” in banks. For example, imagine a bank being able to apply for mortgages on customers’ behalf or save money from their monthly pay cheque.

Starling’s Rees is equally as optimistic. Over the coming years, she expects AI to become a “critical differentiator” in the banking industry — just as mobile banking once was. She says that, to keep ahead of the competition, banks should focus on using the technology for improving financial literacy and accessibility and tackling fraud.

Yuval Samet, CEO and co-founder of AI-powered financial insights platform RiseUp, also expects AI and open banking — banking platforms designed with interoperability in mind — to unlock a major paradigm shift in banking in the future.

He envisages future banking products that do not just provide financial recommendations but put them into action for users, concluding: “Banks that move quickly to integrate AI and open banking within existing experiences will define fintech leadership in the next decade.”