Where people draw the line when an AI agent spends their money

The question. Agentic commerce was being built before anyone had data on how people would accept it. Would people trust an AI agent with their money? What would make them feel safe, and what would make them walk away? Would merchants see agents as an opportunity or a threat to their relationship with customers?

How we ran it. Our AI interviewer spoke with 40 consumers across demographics and 20 e-commerce decision-makers. Each conversation followed up on what the person had actually said: trust, payment preferences, which decisions they would hand over and which they would keep. The VCA team tested the interview and adjusted questions during fieldwork, and changes went live immediately. Screening and interviews ran on our platform end to end.

What people told us. The study mapped where consumers draw the line on letting AI spend for them, and what merchants fear about losing direct contact with customers.

What this shows. This is value data in its most practical form. People describe when an autonomous system may act, when it must ask, and what breaks their trust. It is the same kind of boundary we now collect for post-training.

“The AI moderator handled complex, nuanced topics — like payment security preferences and trust in autonomous transactions — with a level of depth we didn’t expect.”

Artur Bartosiak
Director, Visa Consulting & Analytics, CEE AI & Stablecoin Lead

What came next. Visa made the collaboration ongoing.