Running AI interviews

AI interviews run in the candidate's browser through the same interview room used for practice sessions, with proctoring active for real hiring interviews and inactive for practice.

Scheduling and lanes

How many interviews can run at once for a workspace is governed by interview lanes: a per-company default plus a per-plan cap of one to three concurrent lanes. A queued interview waits for a free lane rather than failing outright, which is the usual explanation when a candidate reports being stuck on a waiting screen at their scheduled time.

Time zones

Every interview time is stored as a UTC instant and rendered twice: to the recruiter in the company's configured time zone, and to the candidate in whatever time zone their browser reports. The confirmation email and calendar invite the candidate receives are generated in their local zone too, so a scheduling mismatch is almost never a bug — it's usually one side reading the invite against the wrong assumed zone.

The interview room

The room itself is a video call with a client-side proctoring layer running alongside it — gaze tracking, face count and audio checks all happen in the browser before anything is flagged. See Proctoring and integrity for what each flag means and how to act on one.