Only the strongest candidates reach your calendar.

A one-way interview records a candidate answering fixed prompts and leaves a recruiter to watch it later. SIA designs the interview structure from your job context first, then conducts it as a conversation: it asks, listens, and follows up when an answer is thin. What comes back is a graded report with evidence attached, not a queue of recordings.

The transcript of what the candidate said. There is no facial analysis, no voice or tone analysis, and no emotion inference. Candidates are evaluated on the substance of their answers against the criteria you approved for the role, and nothing else. Nothing about how someone looks or sounds enters the grade.

SIA generates the interview structure from the job configuration. From there you reorder, delete or rewrite any question by describing what you want: tune it for seniority, add a scenario question, point it at a responsibility that matters. You review the whole structure before a candidate is invited, and the configuration is retained for reference.

Not on its own judgement. An interview grade is an assessment, and every assessment goes to a human for review. What can act without a recruiter is stage automation running on thresholds you set yourself, and any rule can be switched off at any time without deleting it. The AI never makes the call.

Yes. The agent identifies itself as AI and takes the candidate’s consent before the interview starts. The invitation email, the greeting and the closing message are all yours to write, so candidates know what to expect before they join rather than discovering it once the session itself has already begun.

A report card in two panels: the evaluation on the left, the source material on the right. The full transcript with timestamps, the recording linked to the moment in the transcript so you can jump to an answer rather than watching end to end, and the session’s integrity signals. A link opens the criteria and weights behind the grade.

Hundreds of interviews run simultaneously rather than one after another, so a shortlist is interviewed in the time a single panel slot would take. Nobody waits for a calendar to open, which is what usually decides whether your strongest candidate is still available by the time you get round to them.

Either in real time as a candidate reaches the interview stage, or on demand when a recruiter triggers it. The conversation can start within minutes rather than waiting for a batch or a scheduling round, so a candidate who is ready to interview is not held up by your calendar.

It depends on the structure you configure, since you control how many questions there are and how deep they go. The agent follows up when an answer is vague and moves on when it is not, so a candidate who answers well finishes faster than one who needs drawing out.

You decide in advance. Attempts are configurable from one to three, so if you allow more than one, a candidate who drops can start again. The completion window is 3, 7 or 14 days, and that is the deadline reminders work toward. There is no mid-interview pause, which keeps sessions comparable across candidates.

Yes, within limits you set. Configure between one and three reminders, delivered against the completion window you chose. It stops as soon as the candidate starts or the window closes, so nobody is chased past the point where it is useful and nobody is chased after they have already begun.

SIA surfaces tab switching, exits from fullscreen and multi-screen usage, and flags evasive or off-topic answers while nudging the candidate back to the question. These are signals for a recruiter to weigh, not verdicts, and none of them rejects anyone. Each is logged against the transcript moment that triggered it, which makes it reviewable and answerable.

Every candidate is interviewed against the same structure and graded against the same criteria your team approved. Every grade opens beside the transcript passage that produced it, so you can check the reasoning. Criteria are versioned, so the standard applied to a past interview is still on file, and SIA refuses instructions relating to legally protected characteristics.

Yes. The recording sits on the candidate’s report card alongside the transcript, visible to users in your workspace who have access to that role. It is not shared outside your workspace and it is not used to train models. For how long recordings are retained and how deletion works, see the trust centre.