Case · human-in-the-loop

When should an AI agent pull a person in — and hand off without dropping the thread?

A supervision desk for teams of agents — the handoff is the product.

Today agent hand-offs are abrupt — a case lands in a person's lap with little context, or the bot never knows when to let go.

Why now as AI agents take on real support volume, the human hand-off is what keeps trust intact.

Use cases Customer support, contact centres, sales, healthcare intake.

The real load Supervising a team of agents is a new kind of task-switching: you jump between cases mid-thought and each one demands a genuinely complex judgement — approve, correct, escalate, take over. That’s extreme cognitive load carried continuously, and the interfaces for it barely exist yet. Making that load legible — what needs me, what can wait, what the agent already tried — is what this desk is really about.