Using Teamcast

Interview lifecycle

Every interview moves through a fixed set of states, and a human sits at every checkpoint that matters.

An interview is never fully automated end to end. It moves through the states below, in order, and cannot skip a human checkpoint. Every transition is recorded append-only, with the acting user and a reason where relevant — nothing is silently overwritten.

States

StateWhat it means
CREATEDThe candidate has been added to the job. Nothing has been generated yet.
GENERATING_PLANThe AI is building the résumé-informed interview plan.
INFO_NEEDEDThe planner needs more input before it can produce a plan.
PLAN_PENDINGHITL-1: the plan is ready — a recruiter must approve, edit, or send it back.
PLAN_APPROVEDThe plan is locked and the interview is ready to invite.
INVITEDA tokenized join link and OTP have been sent to the candidate.
IN_PROGRESSThe interview is live. It can be monitored, but not scored, in real time.
ASSESSMENT_PENDINGDeterministic scoring runs automatically as soon as the interview completes.
ASSESSMENT_READYHITL-2: a human reviews the coverage assessment and decides advance / finalize / reject.
COMPLETED / REJECTED / CANCELLEDTerminal states — always reached through an explicit human decision.

The three HITL gates

"HITL" means human-in-the-loop: a point where the system stops and waits for a person to act, rather than proceeding on its own.

GateWhat a human decides
HITL-0Role-profile weighting — which competencies matter, and how much, before any candidate exists.
HITL-1Plan approval — the exact questions Maya will ask.
HITL-1.5 / HITL-2Assessment review — coverage is read and a hiring-pipeline decision is made by a person.
Scoring never runs during a live interview. Coverage is computed only after the interview ends, deterministically, from the recorded evidence — never in real time and never influencing Maya mid-conversation.
Was this page helpful?