AI isn’t the story yet
For all the talk of AI as banking’s defining shift, it’s striking how often the same example still comes up: Microsoft Copilot.
Not as a throwaway reference, but as a default. The thing people point to when they want to make AI feel real.
It makes sense. It’s visible. It works. But it’s also telling.
Because when one of the most widely cited examples of AI in banking is still a layer sitting on top of existing systems, it suggests the technology isn’t as embedded – or as mature – as the narrative might suggest.
That gap between ambition and reality ran through much of the conversation.
AI is consistently framed in expansive terms. At Lloyds Banking Group, for example, CIO Ülku Rowe described it as a “superpower” with the potential to reimagine banking. But in practice, most deployments remain focused on internal efficiencies – improving workflows rather than fundamentally reshaping them.
There are clear reasons for this. Embedding AI into core systems is complex, particularly in environments shaped by legacy infrastructure. But the result is an industry making progress at the edges, while the centre moves more slowly.
The same can be said for customer experience.
The shift from products to experiences is now well established. Banks are being judged less on what they offer, and more on how it feels to use. Services are expected to be embedded, intuitive, and responsive – part of a broader financial journey rather than a series of transactions.
At Starling Bank, Group CIO Harriet Rees pointed to a “real shift in UK banking experience.” But she was equally clear on the constraint: money remains deeply personal. However, advanced the technology becomes, banking still needs to feel human.
That balance – between automation and trust – is becoming harder to maintain as systems grow more complex.
At SWIFT, UK & Ireland CEO Adam Bealey described trust as the connective tissue of global finance, particularly as cross-border payments become faster and more interconnected. But greater connectivity doesn’t automatically mean greater cohesion. Without alignment on standards and infrastructure, there’s a risk of creating systems that are more advanced, but also more fragmented.
Fraud is already exposing those cracks.
Described by one panel as the first fully automated criminal economy, it’s evolving rapidly – from synthetic identities to increasingly sophisticated deepfakes. Simone Poon highlighted a 224% rise in synthetic identities and the growing ease with which verification methods can be bypassed. The conclusion is clear: single-layered defences are no longer viable.
But even multi-layered approaches have limits if they’re built in isolation.
Leon Ifayemi made the point directly: the problem is global, but the response is still fragmented. As systems become more complex, accountability becomes harder to define – even as expectations for it increase.
Alongside this, there’s a broader reset happening across fintech.
Dan Dall’Asta pointed out during the VC start up panel to a return to fundamentals: capital efficiency, real problem-solving, and deep domain expertise. Increasingly, the founders best placed to deliver are those building from within the system – not outside it.
And underpinning all of this is resilience.
No longer just an operational concern, it’s now reputational. Outages are visible. They shape trust in real time. That shift is forcing institutions to move from reactive response to proactive design – building systems that are robust from the outset.
Taken together, these conversations point to an industry in transition.
The direction is clear: more intelligent systems, more embedded experiences, more automation. But delivery remains uneven. In many cases, progress is still incremental rather than transformative.
Which leaves a final challenge.
Not just building better systems, but making those improvements visible, understandable, and trustworthy from the outside.
Because the real measure of change won’t be how advanced banking becomes.
It will be whether it actually feels different.