Consumers Trust AI Advice but Resist Letting Agents Take Action

Consumers are broadly open to AI assistance but remain wary of autonomous action, according to a Ping Identity survey of 11,000 people across 12 countries.

The company found that 56 percent were comfortable using AI for support or recommendations. Only 5 percent were comfortable allowing an agent to act with their approval, and 3 percent accepted routine transactions without final approval. Forty-two percent would cap an agent's spending at $100, while 22 percent would prohibit financial transactions entirely.

Identity and representation were central concerns. Seventy-one percent worried that agents could misrepresent their personality or values, and 47 percent said constant oversight was a bigger concern than a single mistake. Only 12 percent fully trusted organizations to manage identity data, down from 17 percent in 2025.

For enterprises, the findings support granular consent, transaction limits and clear audit trails. Agentic identity systems will need to prove not just who the user is, but what authority the agent received, when it was granted and how it can be revoked.