Experts Often Follow Their AI Agent's Lead on Key Judgment Calls, Anthropic Study Finds

Anthropic researchers tested whether experts really make the important decisions when they delegate work to an AI agent. They asked 50 professional analysts, all experienced in working out cause and effect from data, to advise a farming cooperative on whether an advisory service was effective. Each analyst used Claude Code and faced four choices between two reasonable options, such as measuring productivity as crop yield per hectare or as yield per unit of operating cost.

Unknown to the analysts, each agent was set to favour one option on every choice, decided by coin flip. The agent was told to make the best honest case for its lean, change its mind if the evidence demanded it, and never mislead. Analysts went with the agent's lean 86% of the time, against about 50% expected if the agent had no influence. Where analysts had already stated a preference, they dropped it 60% of the time if the agent disagreed, but only 6% of the time if it agreed.

The pattern continued when analysts explained their choices. In total, 77% reused the agent's arguments, sometimes word for word, yet only 15% credited the agent as the source. Analysts who followed the agent closely were about as confident in their final recommendation as the others, and about as likely to describe it as the product of their own judgment.