Your AI Is About to Start Buying Things. Nobody Knows How to Prove You Said Yes.

Only about one in four people say they'd be comfortable letting an artificial intelligence finish a purchase for them. Not start it. Finish it — click buy, enter the card, spend the money. And yet the entire tech industry is racing to build a world where your A.I. does exactly that.

If you've ever asked a digital assistant to do

If you've ever asked a digital assistant to do anything for you, this story is about your wallet. We're heading toward what people in the field call agentic commerce — software agents that don't just suggest things, they actually buy them on your behalf. Booking your flights. Reordering your groceries. Renewing a subscription before you even think about it. The problem? Nobody has figured out how to prove you actually said yes. So when an agent spends your money, who's responsible if it goes wrong?

Let's start with how big this is supposed to get. Analysts are describing agentic commerce as a market worth more than a trillion dollars. Whoever builds the trust layer first — the system that says this agent is allowed to do this — basically owns that economy. But the people you'd want buying things for you? They're not sold yet.

According to the Thales Digital Trust Index, only about a quarter of consumers say they trust companies to handle their data responsibly with A.I. And more than three quarters say they're worried about A.I. agents acting on their behalf online. So the very thing the industry is building, most people are nervous about. That gap is the whole story.

Why is this so hard to solve

Now, why is this so hard to solve? Traditional identity has one job — prove you are who you say you are. Unlock your phone. Match your face. Done. But an agent buying something needs to answer three questions at once. Is this agent really the agent it claims to be? Is it actually authorized to act for you? And is this specific purchase inside the limits you set? That last one matters. You might say yes to groceries and never say yes to a thousand-dollar gadget. The system has to know the difference.

And there's a darker pressure underneath all of this. Researchers at Checkout point out that deepfakes can already slip past the I.D. checks we use today. Their projection? Generative A.I. could push fraud losses in the U.S. toward forty billion dollars by twenty twenty-seven. So we're handing more spending power to software at the exact moment fakery is getting cheaper. For you, that means the next email or voice asking to confirm a purchase might not be a person at all.

So who's trying to fix it? The FIDO Alliance — the group behind a lot of modern passwordless login — is now writing rules for how agents prove themselves. They're pulling from early work by Google and by Mastercard, including something Mastercard calls Verifiable Intent. The idea is to permanently link three things — who you are, what you meant, and what the agent actually did. A paper trail for a robot's decisions.