Most AI hiring tools ship a number and ask you to trust it. "This candidate scored 8.5/10." We refuse to ship that.
A score with no evidence isn't a vouch — it's a guess in a confident voice. The thing we sell isn't a higher score; it's a higher standard for what a score is allowed to be.
If we can't underwrite the claim, we don't make the claim.
What that looks like in the product
Every per-qualifier evaluation in Vouch ships with five fields, every time:
- A title — what we're evaluating
- A rubric outcome — Overwhelmingly / Strongly / Meets / Probably / Inconclusive / Does not meet
- A verbatim evidence quote — pulled from a CV, a transcript, a verified diploma
- A confidence level — High / Medium / Low / Insufficient information
- A comment — the AI's reasoning, in plain language
Click any number. You see what fed it.
The override is the product
Every recommendation Vouch makes is overridable in one click. The override is logged with your reasoning attached. Three weeks from now, when the hiring manager asks why this candidate moved forward, the answer is on the card.
One of these answers survives a debrief. The other doesn't.
The bar going forward
This is the principle we ship every feature against: would I vouch for this in court? If not, we cut it. That's why we don't build AI interviewers. That's why every recommendation has reasoning attached. That's why every override is logged.
A number you can't defend is worse than no number at all.