An agent handoff is the structured transfer of a task between agents, or between an agent and a human, including the context, state, and authority the receiver
A clean handoff transfers the context, state, and authority the receiver needs to continue: conversation history, completed steps, and constraints. Sloppy handoffs force rework and create audit gaps, which is why standardized handoff schemas matter in multi-agent systems.
Because when an AI escalates to a person, that person should inherit a usable summary and the customer should never have to repeat themselves. A poor handoff at that boundary is what users experience as the frustrating 'the bot already asked me this' moment.