AI is the force for change in banking

I’ve just been reading McKinsey’s latest piece about the AI-powered bank. It reads like a polite consulting report, but underneath it is a fairly brutal critique of the entire banking industry because what they’re really saying – without saying it – is that most banks are structurally incapable of benefiting from AI. It not because they lack the technology or budget. It is because they lack the architecture, the data, and the operating model to make AI actually work.

We’ve been here before. Every wave of technology in banking follows the same pattern. First, automation in the 1960s: mainframes, back offices, batch processing. Then digital in the 2000s: apps, APIs, channels layered on top. And now AI, which everyone is desperately trying to plug into systems that were never designed for intelligence.

And that’s the problem.

The McKinsey narrative is all about “rewiring for excellence in customer care,” which sounds benign until you realise what rewiring actually means. It doesn’t mean improving your call centre or deploying a smarter chatbot. It means tearing apart the way the bank works and rebuilding it around data, decisions, and outcomes. That’s a big ask for an industry that still runs on COBOL mainframes and organisational silos.

A good example is the core battleground of customer care, where these pains are most visible. Long wait times, inconsistent answers, endless transfers between departments that don’t talk to each other. Customers don’t experience your org chart. They experience the mess it creates.

AI, in theory, fixes that by acting as a unifying layer. It understands context, accesses data across silos, and resolves issues in real time … in theory. In reality, if your data is fragmented, your processes are broken, and your systems don’t talk to each other, then AI just becomes a faster way of delivering a bad customer experience.

It is why McKinsey keeps hammering the idea that you have to redesign journeys end-to-end. You don’t digitise processes or optimise them; you have to redesign them. This is because most customer journeys in banking were never designed in the first place. They just evolved, organically, through decades of product launches, regulatory patches, and technology upgrades.

So now you have Frankenstein journeys. Half analogue, half digital, stitched together with middleware and hope and then, into this mess, you add AI. No wonder most banks are stuck in pilot mode.

This is where the report gets interesting because it shifts from technology to operating model, and this is the bit most executives underestimate. AI doesn’t just change what you do; it changes how you organise to do it. The traditional model – products here, channels there, operations somewhere else, IT in the basement – simply doesn’t work in an AI world. Instead, you need cross-functional teams aligned around customer journeys, with shared data and embedded AI decisioning. In other words, you stop managing the bank as a collection of silos and start managing it as a series of outcomes.

Sounds obvious, doesn’t it?

It isn’t because it means breaking power structures. It means redistributing control. It means that the contact centre is no longer a cost centre but a strategic intelligence hub. It means that technology is no longer a support function but the core of the business. And it means that data – good, clean, connected data – becomes the most valuable asset you have or, to put it more bluntly, you can’t be smart with dumb data. Most banks are still dumb.

So, when McKinsey talk about AI enabling proactive, predictive customer care, what they’re really describing is something far more radical. A world where the bank anticipates your needs, resolves issues before they arise, and engages you in real-time with contextual, personalised interactions. Not because it has better scripts, but because it has better understanding. That’s not customer service. That’s delegated decision-making.

This is where it gets uncomfortable because, once the bank starts making decisions on your behalf – resolving disputes, managing payments, suggesting actions, even executing them – you’re no longer interacting with a service. You’re interacting with an agent. A system that knows you, learns from you, and increasingly acts for you.

Which is exactly where this is heading.

McKinsey doesn’t quite say it in those terms, but the direction is clear. The endpoint of AI-powered customer care is not a better call centre. It’s the disappearance of the call centre. Issues don’t arise because they’re pre-empted. Queries don’t get asked because they’re already answered. Interactions become invisible.

The paradox is then that, the better you get at customer care, the less the customer notices it.

Of course, getting there is another matter entirely.

The report is optimistic about ROI – cost reduction, improved satisfaction, revenue uplift – but it glosses over the sheer difficulty of transformation. This isn’t a six-month project. It’s a multi-year reinvention of the bank’s core.

Most will struggle. Many will fail. But that’s been true of every major reinvention of banks through technology projects over the last decades as banks will do what banks always do: layer the new on top of the old. Add AI to legacy systems. Build another channel. Launch another pilot. And wonder why nothing really changes.

Meanwhile, a few will take the harder path: they’ll re-architect; they’ll unify data; they’ll redesign journeys; they’ll align their operating model to an AI-first world; and those are the ones who will win. The winners will win not because they have better AI, but because they built a bank that AI can actually run which, when you strip away the consulting language, is the real message here.

AI is not the transformation … it’s the force for change.