What changes for transactions once agents hold payment credentials.
Once an AI agent can hold a payment mandate and complete a purchase on someone's behalf, it's tempting to assume checkout itself transforms. For most businesses, what changes is narrower than that, and what doesn't change is worth stating plainly.
6 min read
What genuinely changes
A homeware retailer hears that agents will soon be able to hold payment credentials and complete purchases directly, and pictures its whole checkout flow being replaced by something unrecognisable. That's the overreach version of this story. The more grounded one: an agent holding a mandate to spend up to £150 on a replacement set of dining chairs, on a real person's behalf, still needs the retailer's own systems to confirm the order, take the payment and settle it, exactly the way a checkout does today. What's new is who's allowed to trigger that flow, not the flow underneath it.
The identity and authorisation step at the front of a purchase gets handled differently: instead of a person typing card details into a form, an agent presents a credential proving it's real and a mandate proving a real person approved this specific spend, up to this specific limit. That's a real, meaningful change to how a transaction starts, and it's genuinely new.
What doesn't change
Everything after that identity check runs on the same rails it always has. The retailer still needs to confirm stock, calculate the real price, process the payment through its existing systems, and settle the transaction the way it settles any other sale. An agent holding a mandate doesn't bypass any of that, it triggers it the same way a person clicking "buy now" does. A retailer picturing a wholly new payment infrastructure to build is picturing the wrong problem.
The part still genuinely unresolved
What isn't yet settled is how disputes get handled when the buyer is an agent acting on someone's behalf rather than the person directly: whose responsibility a mistaken purchase is, how a refund request gets verified as genuinely coming from the person who authorised the original £150 spend. That's not solved yet. It's one of the real open questions the organisations building these standards are actively working through.
Where this shows up for a business
Selfe handles the identity and mandate check at the front of a transaction, so a retailer's own checkout, stock, pricing and settlement, keeps running exactly as it does today, rather than needing to be rebuilt around a new kind of buyer.
Do we need to rebuild our checkout for agent-held payments?
No. The change is at the identity and authorisation step, not the settlement flow, which keeps running on your existing systems.
What happens if an agent gets an order wrong?
That's a genuinely unresolved question across the industry right now, not something anyone can state a settled answer to yet.
The next twelve months in agentic commerce.
Most writing on where AI agents are heading falls into one of two camps: everything changes overnight, or nothing really changes at all. Neither is honest. Some things about agentic commerce are already settled enough to build for. Others genuinely aren't decided yet, and pretending otherwise doesn't help anyone plan.
The standards still being written, and why that's not a reason to wait.
The specifications an AI agent uses to connect to a business, prove who it is, and get authorised to spend are all still being actively revised by the organisations building them. That's genuinely unsettled, not a reason to wait. The work that needs doing doesn't depend on any of it finishing first.
What happens to reviews once agents are the ones reading them.
A star rating tells a person something useful at a glance. It tells an AI agent almost nothing it can act on. As more of a property's reviews get read by agents rather than skimmed by people, what makes a review useful is shifting, and a 4.6 average isn't going to be the part that matters.