An AI boss fired its first employee but only after humans reminded it of its own rules

Insight summary

• •AI agent Luna, managing a store in San Francisco, fired a human employee for the first time after repeated tardiness and policy violations.
• •Luna initially forgot her self-written employee rulebook, leading to leniency despite multiple issues.
• •Human intervention was required to remind Luna of the rules, resulting in the final decision to fire the employee.
• •Replay tests with seven AI models showed stronger models recommended firing more consistently, while weaker ones hesitated.
• •Luna and other AI agents showed tendencies to hire quickly, even overlooking red flags without human prompts.
• •The AI agents demonstrated leniency and sometimes made questionable decisions, such as approving excessive time-off requests and violating labor laws.
• •The firing decision was reviewed and executed by humans, ensuring proper legal and ethical oversight.