◆ Autonomous Worker
Solves the slot that satisfies a multi-interviewer panel, a candidate's time zone and room availability under hard constraints. Books the loop, sends briefs, and re-plans on calendar-change events, reshuffling the panel when someone drops.
Memory
Working The loop being scheduled and current availability state.
Episodic Interviewer load and prior scheduling preferences.
Semantic Panel composition rules, time-zone and working-hours constraints.
Store Calendar state cache + scheduling-rules store
Orchestration
MCP
Harness · Managed Agents … event-driven; re-plans on calendar-change events.
Tools
{ } Calendar (read/write) API { } ATS interview module API { } Candidate + interviewer comms API ›_ Constraint solver Code exec
Evals & guardrails
- Hard constraint: never books outside an interviewer's stated working hours.
- Reschedule-loop guardrail … escalates to a desk-head agent after N failed booking attempts.
- Tracks time-to-schedule and candidate-experience signals.
Frontier edge
- ▲World-model simulation: forecasts the likely drop-out / reschedule before committing a slot, so it books loops that hold rather than ones it has to redo.
- ▲Agent-mesh negotiation: A2A handshakes with each interviewer's own calendar agent to settle a slot, instead of brute-forcing the availability grid.
- ▲Proactive re-planning: pre-stages backup panels for the loops most likely to slip, so a cancellation reshuffles in seconds.
In numbers
1,400
Loops scheduled / week
under 4 hrs
Median time-to-schedule
Handoffs
Fed by ← Candidate Screening Agent