Every payments conference has an AI panel this year. Sometimes more than one. Occasionally several. Most begin with the same question: how will AI transform payments?
After moderating a panel at ACI Worldwide’s Payments Unleashed Conference in London last week, I’m starting to think this maybe the wrong question for 2026.
At this stage in the technology’s evolution, it’s better to ask where is AI already creating measurable value. On this, there was remarkable agreement among the panellists. Fraud.
Whether it was Nvidia talking about Revolut’s Transaction Foundation Model, ACI discussing self-learning fraud models or PayPal describing how AI is improving internal decision making, nobody questioned that AI investment in fraud prevention is producing genuine returns today. Rule-based systems are being replaced with models that continuously learn from new transaction data, improving detection while reducing false positives.
The second area where AI is already delivering results is less obvious but equally important. Payment routing.
Using AI to decide which acquirer to use, when to retry failed authorisations or how to optimise approval rates is producing measurable commercial benefits. Unlike some of the more speculative AI use cases, fraud and routing are straightforward optimisation problems which the industry has been working on for years. Better performance is easily measured and delivers clear cash benefits.
What next? Much of the excitement today around AI centres on agentic commerce. The popular image is one of computers buying from other computers with no human involvement.
Brenden Lane from PayPal argued that’s not what we’re seeing today. Instead, AI is helping consumers decide what to buy rather than buying it entirely on their behalf. Travel and consumer electronics were the standout examples. AI can compare itineraries, explain technical specifications and narrow the choices before the customer makes the final decision. In other words, AI is improving shopping more than it is automating payments.
Erich Litch from ACI agreed. “We automatically jump to the payment,” he said. “There’s a lot of stuff that’s way more valuable in the shopping experience.”
If AI can remove friction from product discovery, supplier selection or travel planning, then the payment simply becomes the final confirmation of a decision that has already been made. The question for payment companies is whether the really changes much in their world. This could just be business as usual and, with some minor modifications, exiting card rails will work just fine.
But investment is certainly growing. Nvidia pointed to rapidly increasing AI budgets and a surge in new application releases. But there was broad agreement that agentic commerce remains at an early stage. Brendan Lane was explicit that it won’t replace traditional commerce this year, while Erich Litch compared today’s adoption levels with online banking in 2000. Interesting, certainly. Mainstream, not yet. And the lesson from 2000 is that the early innovators – remember Egg? – didn’t hit the jackpot.
Any discussion about AI can’t ignore public policy. The USA and Europe are facing in different directions with the UK rather caught in the middle. Rather than calling for sweeping new AI regulation, Lord Chris Holmes argued for technology-neutral, principles-based legislation that can survive changes in technology. It is an attractive idea, although agreeing those principles may prove more harder than writing them down. And there’s an open question whether any regulatory regime can move at the same speed as today’s product roadmaps. Sensible rules today could become dangerous tomorrow.
Less contentiously, Holmes reminded the audience that many current AI projects will fail, not because the technology is inadequate, but because they lack a clear business purpose. This is a common problem when CEOs get excited about a new technology and demand something they can put in a press release.
I don’t think the payments industry struggles to adopt new technology. The challenge is always about identifying where technology genuinely creates value and balancing investment against future returns. It’s looking increasingly like agentic commerce won’t hit prime time in the next year or two. Payments businesses might be best advised to focus investment on fraud and routing until the business case for agentic commerce is clear.
