The whole interview, from brief to evidence.
Tracy AI Interviewer sets up the questions, runs the interview on video, follows up in the person’s field and hands you a result you can check.
Questions from your brief
Tracy drafts questions from the job description or your brief. Edit them, replace them with your own, and set the style from quick to deep and from behavioural to technical.
- CV check adds 3 to 5 questions from each CV and the must-haves
- Skills to evaluate, chosen as tags
- One interview language, chosen at setup
Three ways in
Send a personal invite, publish an apply page, or let your own system ask for an interview and hand the person the link.
- No install for the person being interviewed
- Ready in under 10 seconds, at any hour
- People are told about the recording before it starts
A two-way video interview
Tracy speaks, listens and answers questions about the process. The interview is recorded on both sides, with a transcript timed to the recording.
- Rejoin the same session within 10 minutes
- A microphone check before the first question
- A reminder goes out after a day for unfinished interviews
How much faster were reads, and what did it cost you on writes?
She waits, then asks the next question
Tracy waits for the end of a thought, about half a second on simple questions and up to about two seconds on hard technical ones, and checks the thought is complete. Then she asks about what was actually said.
- Each follow-up depends on the answer before it
- A vague answer gets a sharper question
- No cutting people off mid-sentence
A coding task, then the trade-offs
For technical roles Tracy can introduce a coding task and probe the trade-offs behind the answer: memory, failure modes, what breaks at scale.
- Questions about the code that was written
- Follow-ups that test the reasoning, not recall
An assessment for every question
After the interview Tracy writes an assessment for each question with the reasoning behind it, plus an overall assessment. Warm-ups, notice period and salary questions are left out.
- Every point links to the moment in the recording
- The full transcript beside the video
- Tracy does not reject anyone. A person decides.
Index for the ledger table
Chose the index from measured read latency, and named the write cost without being asked.
A migration that went wrong
Owned the rollback. Gave the order of steps and what she changed in the runbook after.
Retries from a payment provider
Idempotency keys on write. Did not cover what happens when the key store is down.
Language skills, when they matter
For roles where spoken language matters, a separate evaluation covers fluency, coherence, grammar, pronunciation and vocabulary on a 0 to 100 scale.
- Shown as a radar beside the transcript
Signals with their evidence
Tracy captures signals during the interview: the face and where the person is looking, response timing, tab switching, leaving fullscreen, an external screen and repeated requests for a question. They are assessed together afterwards and the evidence goes to the reviewer.
- Signals, not a verdict
- Marked on the recording timeline
- Nothing is decided during the interview
API and MCP
Tracy runs as an MCP server, so a team can set up interviews, invite people and read results from Claude, ChatGPT, Mistral or Slack. A customer’s own system can request an interview through the API.
- Nothing is sent until you approve the preview
- Pair it with your ATS’s chat connector, such as Workable’s
- Enterprise teams can ask for a customised setup
See it in your kind of interview.
Let Tracy take the next interview.
Run your first 3 interviews free, or book a demo and watch Tracy work through your own questions.
Your first 3 results are free. No card needed.