Andy Sack spent a full day managing a complex writing project through Claude Cowork's dispatch feature, directing his AI agent from his phone between coffee meetings. The result: roughly ten times the output he could have produced working alone. Then at step five of six, the agent told him his files didn't exist.

This mini episode captures the full arc of working with AI agents: the invigoration of real-time delegation, the gut punch when things break, and the 90-minute recovery that still left Andy ahead of where he'd have been without agents. Adam Brotman contextualizes the experience, explaining how agents differ from chatbots and why the failures are instructive, not just frustrating.

The conversation also covers Andy's rapid skill-building, going from 5 saved skills to 28 in two weeks, and what that signals about how individuals and organizations will work with AI agents going forward.