In a recent article, the author shares their experience of automating their workflow using ChatGPT agents, highlighting a critical failure they encountered. Initially, the agents were generating incorrect information, such as wrong deadlines and imaginary stakeholders. However, by implementing a simple yet effective constraint—forcing the agent to admit when it doesn't know something—they transformed the entire system. This insight is significant as it emphasizes the importance of clear communication in AI, ensuring that automation tools can operate effectively without making unfounded assumptions.