An AI agent that runs the admin for a small business keeps working after the owner goes to sleep. It messages clients, books meetings and follows up, and some of that should never happen without her. Which actions should wait for her? What does the agent do while they wait? And when she replies in the morning, how do you know what she said yes to? The talk answers all three, and shows how trust builds up until she doesn't have to be asked about the same thing every night. You'll leave with a way to decide which actions your agent can take alone, and practical advice on implementing it in a way that fits your use case.
Principal Engineer @ Honeybook