Three parts. Only one of them scores.

Xpert reads a person the way a good manager would. It measures what can be measured properly, then spends the rest of its time understanding the detail. The measuring and the understanding are kept apart on purpose.

A test on its own is a score with no context.

A resume parser is keywords with no person.

An AI interview with nothing validated behind it is just a conversation.

Xpert is all three, and only the validated part is allowed to score.

What each part does

The three parts are not three versions of the same thing. They answer different questions, and each one is allowed to touch a different part of the result.

01 THE MEASURE

The assessment

Thirty eight questions, built on Big Five and RIASEC. Decades of published validation behind them, and the same questions for everybody, in the same order.

WHAT IT PRODUCES

THE ONLY THING THAT SCORES

02 THE CONTEXT

Your work history

A resume, a Word file, a Google Doc, a photo of a printed page, or typed in by hand. Whatever somebody already has, we read it, because asking a person to go away and convert a file is asking them not to come back.

WHAT IT PRODUCES

NEVER REACHES A SCORE

03 THE NUANCE

The interview

A spoken conversation that already knows where somebody has worked, so it asks what that job was actually like rather than asking somebody to sum themselves up. Typed instead, for anyone who needs it.

WHAT IT PRODUCES

ENRICHES, NEVER SCORES

The order matters

Each part earns its place by making the next one better, and the scored part never moves.

The assessment runs first and stands alone. Its result does not depend on anything a person uploads or says afterwards.

The work history makes the interview specific. An interview that opens with "tell us about your work" gets a shrug. One that knows where you have worked asks a real question.

The interview is where results stop sounding like a report. It adds depth and language, not points.

The guardrail

Everything above only holds if one rule holds. It is built into the structure rather than promised in a policy.

Nuance in. Identity out.

Protected characteristics cannot reach the engine that produces a score. Not as inputs, not as features, not as tiebreaks. A test fails the build if they ever do.

Never scored: Nothing said in an interview or written on a resume moves a trait, a fit label or a ranking.
Never weighted: Anything a person chooses to record about themselves is held for them, not fed to the matching.
Never inferred: No component guesses age, race, sex or origin from a name, an accent or a turn of phrase.

Built to stand up to AEDT bias audits: NYC Local Law 144, Illinois AIVIA, Colorado SB205.

Download the one page version