Every UK bank says it’s ‘exploring AI’. But in reality, most mean they’re ‘waiting for permission’.
Recent research shows that barely half of UK banks have actually deployed AI. Instead of moving forward, many are tiptoeing around lagging regulations, risk-averse boardrooms, and legacy systems that make change challenging.
Meanwhile, more agile fintechs and digital challengers are folding AI into the fabric of their products – delivering sharper insights, a smoother customer experience and leaner compliance and operations.
As a result, the gap between AI leaders and laggards is widening. And the uncomfortable question is becoming unavoidable: will the incumbents catch up?
So how do UK banks break out of this AI quagmire? They must admit the problem isn’t that the FCA is blocking progress. It’s that banks are treating the FCA’s outcomes-led approach as a reason to pause.
Hiding behind the regulator
The FCA is playing the long game. It is staying outcomes-focused, rather than reinventing the rulebook in response to the AI boom. And while this does have the potential to create ambiguity, it does not need to be prohibitive.
Yet too many banks are playing chicken with the regulator, using uncertainty as justification for inaction. While Westminster debates AI strategy and MPs criticise the FCA’s “wait and see” approach, bank boards are quietly relieved at the excuse to defer decisions.
While the UK is in limbo, banks are looking to the EU AI Act to see what their regulatory futures may hold. Specifically, this means setting up granular reporting for AI systems and strict governance on “high-risk” systems like credit scoring and AI-driven investment.
The result is predictable. Banks are working on pilots that never scale, and “safe” front-end generative AI (often chatbots) that deliver limited impact. These low-impact projects are easy to demo, but hard to scale, and rarely move banks’ balance sheets meaningfully.
The real opportunity is inside the bank
AI’s greatest value for banks is not conversational – it’s operational.
By focusing on the safe options and avoiding AI use across potentially higher-risk systems, banks are failing to tap into all-important efficiency gains. The higher cost and more complex processes sit deep within the organisation: reconciliation and exception handling, interest accrual, payments monitoring and investigations, financial crime case triage, batch processing, credit assessment, or risk analytics.
These are not flashy use cases. They are high-volume, rules-heavy processes where manual effort dominates. They are also areas where AI can deliver meaningful efficiency gains without fundamentally rewriting consumer-facing risk models.
Take AML triage. Embedded properly, AI can help to prioritise cases, reduce false positives and free investigators to focus on high-risk cases.
Done well, these applications reduce queues, improve consistency and cut operational drag. This creates space to improve products and deliver personalisation, without ballooning costs.
Connecting the serious stuff to AI
Of course, the risks are real. Poor data, inherited bias and weak oversight can create serious problems at scale. Regulators will eventually tighten scrutiny – and banks will be expected to explain their decisions and use of data in detail.
The principles for any future guardrails won’t be new. Looking at regulations like the FCA’s Consumer Duty, banks can ensure that they will follow the regulator’s age-old principles of fostering accountability, control, and producing evidence throughout systems.
If the UK banking industry wants to lead on AI, the banks themselves must address these risks head-on, building the right technology foundations and safely connecting critical systems to AI today.
Ultimately, banks must be able to answer: why was this decision made, and can you prove how the model arrived at it? But, without trusted, timely data and systems that can provide it, this is frankly impossible.
Banks that want to move forward must take the initiative, ensuring their most valuable systems have strong guardrails that will make AI models auditable, secure, and cost-effective to run. This includes standardising processes to reduce exceptions, and putting governance in place that treats models as regulated components.
Let’s call it instinctive banking. The instinctive organisations will have broken down internal silos and embraced external partnerships, potentially even with their competitors. These organisations can easily share data, intellectual property, skills, and more for a seamless user experience that benefits their customers
Harnessing data and advanced technologies, instinctive organisations gain real‑time actionable insight and fast, accurate prediction capabilities so they can anticipate customer needs, business opportunities, and risks. This also gives banks the perfect platform to improve customer insights and streamline back-office functions, so they can deliver serious efficiency gains.
It’s time to steer the ship
AI is a once-in-a-generation opportunity for banks of all sizes. But if the UK wants to lead in AI-driven finance, banks can’t be passengers; they need to steer the ship.
That means modernising their core banking infrastructure, fixing data fragmentation, improving data quality, and embedding AI where it can deliver real operational gains – not just cosmetic front-end upgrades.
Banks that act now will build resilience, efficiency and customer stickiness for the next decade. The ones that wait may find the market has already moved on.
