Why are we banking like it’s 1999?

The Bank of England has released its latest report on artificial intelligence in financial services. The numbers are impressive. The reality is less so. Banks are embracing AI at extraordinary speed, but are they transforming banking or just making yesterday’s banking cheaper?

The Bank of England and Financial Conduct Authority have just published their latest survey of artificial intelligence in UK financial services, and the headline number is remarkable: 86% of financial firms now use AI, compared with 75% in 2024.

That’s almost everyone.

Fantastic. The revolution has arrived! The machines are taking over! The future of banking is here …

… except there’s a problem.

When you dig beneath the headlines, you discover that much of this extraordinary AI revolution involves helping bankers write emails, summarise documents, produce code and automate administration.

In other words, we’ve invented technology that could fundamentally reinvent money, banking and financial services … and we’re using it to write meeting notes.

Welcome to the future.

The Bank of England’s 2026 AI report contains some fascinating statistics.

AI is now mainstream. Generative AI accounts for 56% of reported AI systems. Operations and IT represent the largest concentration of use, accounting for 31% of systems and 80% of firms report deploying AI agents.

Four out of five firms.

You would imagine that meant autonomous digital financial assistants managing our money, moving our investments, negotiating mortgages and organising our financial lives but no, not quite. Most of these agents are supporting decisions rather than making them.

Enterprise productivity and coding assistants dominate, while banks remain cautious about giving autonomous agents control over payments, trading and customer-facing decisions.

The industry has embraced artificial intelligence while keeping much of the intelligence away from the customer.

It’s an extraordinary contradiction.

Banks are racing to adopt the technology that threatens to reinvent their business model, but much of their investment is focused on preserving the old business model rather than rethinking it.

This is the core thesis of Intelligent Bank. You cannot just go from mainframes to digital to AI. You need to rethink, rebuild and reboot. Hardly any banks are doing this. They’re making the existing machine more efficient rather than asking whether the machine still needs to exist in its current form and that’s the real story here.

Here’s my favourite statistic from the report: just 2% of respondents have deployed AI agents with broad autonomy meanwhile, even including planned deployments over the next twelve months, no more than 9% expect agents to execute payments, make trading decisions or approve customer outcomes.

This tells us something important about the gap between artificial intelligence and artificial transformation. Banks love AI when it reduces costs yet they’re far less enthusiastic when it starts making decisions. That’s a hypocrisy I blogged about the other day but there are good reasons for caution. Nobody wants an autonomous agent hallucinating a mortgage approval, sending £10 million to the wrong account or making a thousand disastrous trades before anyone notices.

The thing is that the issue isn’t whether banks should abandon controls. It’s whether they can redesign those controls for a world in which machines increasingly act on behalf of people with autonomy because the next generation of banking isn’t about asking an AI chatbot what your balance is. It’s about telling your financial agent what you want to achieve and letting it organise your money to make it happen.

My money has an agent. Does my bank? That’s the question financial institutions should be asking.

There’s another fascinating finding buried in the report which is that third-party providers now account for 39% of reported AI systems, and 53% of those classified as highly material. More strikingly, the three largest cloud and computing providers account for 79% of reported provider relationships.

Read that again.

An industry that has spent decades worrying about systemic risk, concentration and institutions being too big to fail is building its AI future on infrastructure concentrated among a handful of technology companies.

We’ve spent years regulating banks because they’re critical to the economy. Now we’re handing more of the technological foundations to companies that aren’t banks. What happens when one of those providers fails? What happens when its systems are compromised? What happens when its commercial interests conflict with those of the financial institutions depending on it? And who really controls the intelligence powering tomorrow’s financial system? The banks? The regulators? The technology giants? Or nobody?

This came up as the big question when Google Cloud failed the other day, but it is more than just a technology question. It’s a question of power, dependency and sovereignty.

If data is air, AI is the purifier and the people who own the purifiers increasingly control the economy (the theme of my latest book: Who owns the Oxygen?).

The report reveals another uncomfortable contradiction in that an impressive 94% of respondents have firmwide AI policies, principles or standards.

Wonderful. The compliance departments have been busy … but then only around half report ongoing monitoring, leakage testing and periodic review or recertification controls.

So we have an industry that’s exceptionally good at approving AI before deployment, but far less consistent at monitoring what happens afterwards. That’s rather like approving a self-driving car because it passed its driving test and then deciding there’s no need to check how it behaves on the motorway.

AI systems evolve. Models change. Vendors update their software. New vulnerabilities emerge. But the risks don’t remain static.

The report finds that 57% of respondents already regard AI-amplified cyber risk as a high or severe systemic concern. That rises to 80% when looking one to three years ahead.

In other words, the industry is accelerating into an AI-powered future while acknowledging that the dangers are growing.

The question is whether governance can move as quickly as the technology.

Perhaps the most revealing findings in the Bank’s report are concerns about the barriers to adoption.

Among the leading non-regulatory constraints are integrating AI into existing business processes and dealing with legacy technology.

There’s the elephant in the room.

You can buy the smartest AI in the world, but if you’re connecting it to fragmented databases, ageing core banking systems and decades of organisational complexity, you’re putting a Ferrari engine into a horse-drawn carriage.

I think I’ve made this point too many times before.

I guess my takeaway from the Bank’s report is the final irony: AI could make banks better while making banking obsolete.

This is the challenge that should keep every bank chief executive awake at night.

The Bank of England’s report shows an industry embracing AI enthusiastically. That is encouraging but it also reveals a sector largely concentrating on productivity, efficiency, risk management and operational improvements. Those are valuable achievements. They save money, improve services and reduce friction.

Critically, however, they don’t guarantee survival.

The bigger threat comes from organisations that don’t have to preserve the existing banking model. Technology platforms, fintech companies and AI-native financial providers can start with a different question. Instead of asking how AI can make a bank more efficient, they ask why the customer needs to interact with a bank at all. Why open a banking app when your financial agent can handle the transaction? Why search for a mortgage when your agent can negotiate with lenders? Why move money manually when intelligent systems can manage liquidity continuously?

Banks remain essential to regulated financial intermediation, safeguarding deposits and managing risk but their role in the customer’s daily financial experience is up for grabs and that changes everything. The great danger for incumbent banks isn’t that they fail to adopt AI. It’s that they adopt AI enthusiastically, successfully and profitably … only to discover that someone else has used it to reinvent their industry.

The Bank of England tells us that 86% of financial firms use AI.

Impressive.

Now I’d like to know how many are using it to reinvent banking rather than just make the past processes more intelligent because there’s a world of difference between an AI-powered bank and a bank built for an AI-powered world.

An AI-native bank.

Source: Bank of England and Financial Conduct Authority, Artificial intelligence in UK financial services – 2026, published October 26, 2026