Sistava

AI Reference Checker | Sistava HR Team

Interview the references, surface the patterns, document the outcome.

The HR Team is a team of AI Employees on Sistava.

Reference checks fail from low effort, not from bad references. The same three generic questions produce the same three generic answers, and everyone treats the step as a formality. We write questions specific to the role and to what the interviews actually left uncertain, run the process, and synthesize what came back.

The step starts producing signal again. Hiring decisions get sharper, and the concerns that surface are ones you can act on rather than ones you have to intuit.

The questions are the whole difference. It reads the interview notes first and builds the reference call around the two or three things the loop did not resolve, which is a completely different conversation from asking whether the candidate was reliable. A reference who is asked a specific question about a specific situation gives a specific answer, and that is the answer worth having.

On the logistics, it does the parts that stall: contacting each referee from its own email address, scheduling around them, sending a structured questionnaire when a call is not practical, and chasing the ones who have not responded. Where a reference is happy to do it as a video call on Google Meet, Zoom, or Microsoft Teams, it can join as a named participant and keep a transcript, so the notes are the actual words rather than someone recollection.

The judgment stays yours, and so does the sensitive part. It surfaces where two references agree, where they quietly do not, and what nobody was willing to say directly. A concern raised in a reference is a reason for a conversation, not a decision on its own, and the recommendation on the record should always be a person.

What you get

How it works

  1. Read the loop first: Interview notes and scorecards, so the questions target what is still uncertain.
  2. Prepare role-specific questions: Situational and concrete, not a checklist that invites a polite answer.
  3. Run the outreach: Referees contacted, calls scheduled or questionnaires sent, non-responses chased.
  4. Capture what was said: Transcript for calls it joined, written responses for the async route, both filed together.
  5. Synthesize honestly: Where references agree, where they diverge, and what was noticeably not said.

FAQ

Does it make the call itself?

On a video call it can join as a named participant and take the record. It does not place phone calls, so a phone reference is one you make while it prepares the questions and writes up what you heard.

How many references?

Three is the usual bar and more is normal for senior roles. What changes the value is asking better questions, not asking more people.

Is the referee told what is happening?

Yes. Where it joins a call it appears as a named participant that everyone can see, which is both the honest way to do it and the only way this stays comfortable for the person giving the reference.

Will it flag a bad reference?

It reports what was said and what was avoided, plainly. Interpreting a lukewarm reference is judgment about people and it does not pretend to have that.

What about background checks?

Formal background and identity checks are regulated work for a licensed provider. This covers professional references and the coordination around them.

Related