vibencode
Case study

AgentR

Screening, structured evaluation and interview logistics, built so two candidates for one role meet the same written definition.

Client AgentR Sector Recruitment
The shape of it The sentence it arrived as
What we heard

“We are rejecting people we should be hiring, and we cannot tell you which.”

What it became

Screening that reads what somebody has done, evaluation scored against a written definition of the role, and the scheduling taken off the coordinator.

What we found The problem under the stated one

Keyword matching does not fail loudly. That is the whole problem with it.

It returns a shortlist, the shortlist looks reasonable, and the candidates it dropped are invisible to everyone involved, so the system reports success at exactly the rate at which it is failing.

Keywords also select for having written a CV in the industry’s dialect, which stands in for a particular career path. And then the part everyone underrates: calendar tennis loses the candidates who had other offers, which arrives on the report as a sourcing problem and is not one.

The system reports success at exactly the rate at which it is failing.

A keyword shortlist always looks reasonable, and the candidates it dropped are invisible to everybody.
What we built In the order it happened
  1. 01Screening that reads the career (what was built, at what scale, what changed while the person was there) rather than scoring terms.
  2. 02Evaluation against a written definition of the role, applied identically to everyone, with the reasoning recorded rather than a bare score.
  3. 03Structured responses to behavioural and situational questions, so two candidates can be compared on the same thing.
  4. 04Scheduling, coordination and feedback collection taken off the coordinator and run to completion.
  5. 05Reporting on the interviewers as well as the funnel, because a panel that rates everybody highly is a problem the funnel cannot show you.
Ships with
  • Reasoning-based screeningReads the career rather than the keyword density, and records why.
  • Structured evaluationOne written definition of the role, applied the same way to everyone.
  • Interview operationsScheduling, coordination and feedback, run to completion.
What it does now In production

Applications are read rather than matched, candidates for one role meet the same written definition, and scheduling no longer costs a week.

Because the reasoning is recorded rather than reduced to a score, a rejection can be reviewed, which is the only way a screen ever improves.

Taken over
  • First-pass CV review
  • Scheduling and panel coordination
  • Collecting feedback after the fact

This one sits across Agents and orchestration and Applied research. Most engagements touch two.

Other engagements

Is yours the same shape as this one?

Book a 30-minute call

Describe it in your own words. We will tell you where it lands.