We’re closer than ever to a world where your AI agent makes a payment, your bank’s agent investigates the dispute, and the merchant’s agent handles the chargeback, all without a person touching a keyboard. This isn’t a far-off vision; it’s already on product roadmaps. The technology will be ready. The question is whether we’ll have earned enough trust for anyone to use it.
Trust Is the Real Bottleneck
Trust is the foundation of the AI era because technology only scales as far as it’s trusted. You can build the most capable agent in the world, but if a customer doesn’t trust it to protect their money and fix things when they go wrong, they’ll never hand it their wallet.
Trust is built in the hardest moments: when there’s fraud, when an agent makes a mistake, when a charge has to be disputed. Customers forget the thousand transactions that clear without a hitch and remember the one time they were defrauded, and whether their bank stood behind them. In a world run by agents, these moments arrive faster and more often, and each becomes a chance to earn a customer for life or lose them for good.
Trust is also fragile, taking years to build and a day to lose, which is why institutions that treat fraud and disputes as a cost to squeeze, rather than a chance to prove themselves, lose the customers they can’t afford to lose.
The Threat Side of the Equation
There’s an uncomfortable truth on the other side of all this: the same tools making commerce easier are making fraud easier too. It’s never been cheaper to run a sophisticated scam because the people attacking us have the same AI we do, and use it on schemes that once demanded real skill. We have to be right every time. They only have to be right once.
The regulatory ground is shifting, too. For years, the blame for someone being tricked into wiring money to a fraudster fell on the sending bank, but regulators are now looking harder at the receiving institution, the one that let a criminal open the account. That makes truly knowing your customers a frontline defense rather than a compliance checkbox, and leaves you responsible not just for your own customer but for what they might do to someone else’s.
Data Is How You Build Trust
So how do you do all this at once—building trust, outpacing a faster and cheaper enemy, keeping up with the rules? For me, it comes down to data.
Almost nobody taps the mountain of data already in the fraud ecosystem, because it stays trapped in systems that don’t talk to one another. The dispute team can’t see what the fraud team sees, the fraud team never learns how disputes resolve, and a warning about a new scam at one bank rarely reaches the bank it will hit next week. If my neighbor’s house were broken into, I’d want to know, because I might be next, and yet most banks have no way of knowing.
Good intelligence warns you about threats before they reach your book and lets you automate with confidence, since automating on bad data just means making bad decisions faster. If you aren’t pulling in every piece of data you can to resolve a dispute, you’re leaving yourself exposed.
My advice: gather as much data as you can, as early as you can, and use it to tell your customer what happened on their account, because being straight with people is itself a form of trust. Then feed everything you learn from that dispute back into how you fight fraud, so the next attack is one you’ve seen before. A dispute is far more than the end of a transaction; it’s the clearest look you’ll ever get at how fraud is moving through your institution, and yet most companies throw that insight away.
What the Winners Will Do
Agentic commerce is coming whether we’re ready or not, and the institutions that thrive won’t be the ones with the flashiest agents but the ones that turn their data into trust, making fraud and disputes the place that trust is won. In the AI era, that’s no longer a back-office function. It’s the foundation everything else is built upon.
