Mastercard Signals Report: Agentic Commerce and AI Trust

This reminds me of when I was getting myself up to speed on crypto and blockchain. IBM built something with Maersk a few years back, similar idea but with blockchain. They wanted to take human out of the loop because they wanted to cut the errors and speed everything up. It actually worked. The delays went down and it was designed to cut fraud too, which was fantastic. Sadly it got shut down, and this wasn't because the tech didn't work. It's because the shipping companies didn't trust putting that much control in a system led by one of their competitors. I think that's the real lesson for agentic commerce too. You can build the smartest, most accurate system in the world. If people don't trust it, they won't use it. Fewer errors isn't the same as more trust. You still need a human somewhere in that loop, not because the AI isn't capable, but because trust has to be built, not just proven on paper. AI knows. It hasn’t lived it - KHS

Thanks for sharing Ann.. we are in dynamically shifting times Agentic commerce changes the authorization model. Traditional authorization is Person to Payment Now with Agentic AI we will authorize: Person to Agent to Tool to Merchant then to Payment.This means every hop is now a delegated authority boundary. One thing that stood out to me is that agentic commerce fundamentally changes where trust is established. In traditional commerce, trust is largely established at the point of payment. In agentic commerce, trust has to exist much earlier, when a customer delegates authority to an agent to discover products, negotiate options, retrieve information, invoke services, and then commit to a purchase. That shifts the engineering challenge as payments now become one checkpoint in a much larger transaction. Our challenge now is ensuring every decision leading to that payment remained within the authority originally delegated, while providing evidence that it did..

Spot on. Intent and guardrails matter, but the real test is what the agent can actually reach once it's already acting, that's where trust holds up or breaks down.

Agentic commerce will change much more than how people pay. It will shift advantage towards the businesses that can build trust, make better decisions and create products that work seamlessly when AI agents are acting on the customer’s behalf.

“As AI agents gain the ability to act on our behalf, trust and security can’t be an afterthought. Intelligence without accountability creates risk.”

The interesting shift is liability, not capability. Once an agent can complete a transaction end to end, who eats the loss on a bad decision becomes the real product question, not just the fraud detection layer.