Category: Research · Source: Fortune · Published: 2026-11-05T12:00:00+00:00

AI agents went shopping for TVs and laptops—and messed up nearly 50% of the time, study shows

Summary

A first-of-its-kind study published in Fortune evaluates the reliability of AI agents in executing e-commerce tasks, specifically purchasing TVs and laptops. The experiment found that agents, including Gemini and Muse, failed in nearly 50% of attempts, highlighting significant gaps in autonomous transactional accuracy. This research serves as a benchmark for the current limitations of agent-to-agent commerce rails, indicating that while the technology is advancing, it is not yet robust enough for unsupervised high-stakes financial transactions. The findings underscore the need for improved error handling and verification protocols before widespread adoption in retail environments.

Key impact

Study reveals 50% failure rate in agent shopping tasks, highlighting reliability gaps in current AgentFi commerce implementations.

Read the full story on Fortune