Stanford: 89% enterprise agents fail

- Social summaries of the Stanford AI Index say roughly 89% of enterprise 'agents' fail in production because of reliability, data miscues, and integration gaps. (x.com) - Recommended mitigations include audits, action caps, and stricter governance to prevent misexecution and uncontrolled agent actions. (x.com) - That failure rate helps explain why large rollouts like Accenture’s Copilot to 743,000 employees require heavy guardrails and change management. (x.com) (x.com)