How willing are people to let AI control their finances?

Public First's global survey — polling more than 18,000 respondents across 15 countries, eight advanced economies and seven emerging ones — reports that trust in AI agents has a hard edge at financial control. "Just 29% of people said they would be comfortable giving an AI agent access to their bank account," the report states. That figure sits well below the overall optimism the same survey finds elsewhere: 45% of respondents believe AI will change the world significantly, and 46% say it already has. The gap between general enthusiasm and willingness to hand over one specific, high-stakes permission — bank access — is the report's clearest evidence that agentic delegation faces a trust ceiling distinct from attitudes toward AI in general.

The report frames this alongside its broader argument that AI's economic upside is not automatic: "around half (47%) of the potential economic value over the next five years depends on the choices we make — how fast we adopt AI, and how prepared we are to use it in more sensitive areas." Banking access is exactly this kind of sensitive-area choice. The survey's occupational breakdown — roughly 45% of occupations where AI stays purely complementary and 36% where it substitutes for the highest-value tasks but humans are still needed for the rest — describes a similar pattern of partial, bounded delegation rather than full handover, which the bank-account figure extends from work tasks to personal financial control.

This bounded-delegation finding complicates the augmentation narrative in Does AI adoption follow wealth and mature over time?, which describes usage deepening toward fuller task handover as adoption matures — Public First's trust data suggests that progression has a ceiling for financially sensitive permissions specifically, not just a general comfort level that rises with exposure. It also echoes the expectation gap in Do executives and employees agree on AI's job impact?: both surveys find a split between how confidently institutions project AI's reach and how far the people on the receiving end are actually willing to go.

The report does not break the 29% figure down by country, age, or prior AI use, so it is not possible to tell from this excerpt whether the trust ceiling is falling, rising, or stable, nor whether it varies between the advanced and emerging-market respondents surveyed. It also does not establish why people decline — security concerns, loss-of-control concerns, or unfamiliarity are all consistent with the number — only that the refusal rate is high. The implication the evidence supports, at this strength, is that agentic-commerce products asking for financial permissions should expect resistance on the order of 70% even where general AI sentiment is positive, and should treat trust in a specific permission as separate from trust in AI broadly.