A rubric-first ATS starts from your hiring rubric (the weighted criteria that define a great hire for the role) and scores every candidate against it. In Rubrily, the AI interview and the CV evaluation each score every criterion 0–10 with a written justification, so every ranking traces back to evidence instead of a black-box match score.
We coined the term. There was no word for the distinction we kept having to explain: plenty of systems have AI, few can show the rubric behind a score. So this page is the term's definition, and the test any vendor claiming it should be held to. Including us.
Why "AI ATS" stopped meaning anything
Every applicant tracking system now claims AI. Vendor positioning has converged on the same vocabulary (AI-powered recruiting software, screening agents, "AI at the core") until the label stopped carrying information. Reading ten product pages tells you that ten companies have shipped AI features; it tells you almost nothing about how any of them scores a candidate.
The question that separates the tools was never whether there's AI in the building. It's what the AI scores against, and whether it shows its reasoning. One system can run a state-of-the-art model and still produce an unexplained match percentage against criteria nobody defined. Another can use the same model to score your criteria, at your weights, with written evidence behind every number. The category label covers both (what an AI ATS is remains a real and useful distinction from traditional workflow systems), but inside that category, the scoring architectures have almost nothing in common. Rubric-first names the architecture, not the ingredient.
The architecture that earns the name
Rubric-first is a chain, and the order matters:
- The rubric is defined once, before screening starts. Weighted criteria, from Must Have to Good to Have tiers, that encode what a great hire looks like for this role. (Starting from a blank page? Here's how to write a hiring rubric.)
- The same rubric drives both evaluations. The async AI interview and the AI CV evaluation score against one set of criteria, not an interview model and a CV model each holding private opinions about what matters.
- Every criterion gets a 0–10 score with a written justification that cites the evidence: which answer in the interview, what in the CV.
- Missing evidence returns "Cannot evaluate," never a fabricated number. And an unscored criterion is excluded from the weighted math rather than counted as zero, so the system's honesty doesn't quietly penalize the candidate.
- Everything rolls up into one Fit Score that decomposes back down, from the ranking number to component scores to criteria to justifications to the transcript and CV they cite. That roll-up is deliberate arithmetic, not another model call: how the Fit Score is computed.
Remove any link and the chain stops being rubric-first. A score without a rubric is a black box. A rubric without justifications is a checklist. A justification that can't be traced to evidence is decoration.
What rubric-first is not
It is not anti-AI. The AI does all of the reading, interviewing, and scoring. That is precisely the labor being automated. The rubric decides what counts; the AI works out how each candidate measures against it. Rubric-first is a statement about who sets the criteria and how the number is earned, not a preference for doing less with the machine.
It is not the same as structured hiring. Structured hiring is a process methodology: consistent stages, planned interviews, defined debriefs. Rubric-first is a scoring architecture: what the machine evaluates against and how it justifies the number. You can run structured hiring on any ATS, but a score is only rubric-first if the rubric produced it.
It is not a keyword filter with extra steps. A keyword filter counts vocabulary; a rubric-first evaluation reads meaning against the criterion and explains what it found (the full comparison is in rubric scoring vs keyword filtering). A candidate who "led the migration off a monolith" scores on architecture experience whether or not their CV contains the word you would have searched for.
The test (use it on any vendor, including us)
Four questions, askable in any demo:
- Can I see the rubric behind any score? Not the marketing page about AI, but the actual criteria and weights this candidate was scored against.
- Does every score carry a written justification citing evidence? Ask for the reasoning behind one specific criterion score, then check that it references something real.
- What happens when evidence is missing? The right answer is an honest "Cannot evaluate." The wrong answer is a number anyway.
- Can a score be decomposed to the transcript or CV it came from? From the ranking number down to the source material, without leaving the report.
If the answer to any of the four is no, the product may be AI-assisted. It isn't rubric-first. These four questions are the hiring-specific edge of explainable AI screening: the general property that AI scores can be checked, applied to the specific machinery of an ATS.
FAQ
Is a rubric-first ATS the same as an AI ATS? No. One contains the other. An AI ATS tells you that AI does the screening work: interviews, CV evaluation, scoring. Rubric-first tells you what that AI scores against and that every score explains itself. Rubric-first is the subset of AI ATSs whose scoring is auditable; every rubric-first system is an AI ATS, and the reverse is not close to true.
Is rubric-first the same as structured hiring? No. Structured hiring is a process methodology: consistent stages and interview plans for every candidate. Rubric-first is a scoring architecture where the machine evaluates against your weighted rubric and justifies every number it produces. The two compose well: structured hiring decides what happens when, and rubric-first makes the scores generated along the way traceable.
Do I have to write the rubric myself? Yes, and that's the point. The rubric is yours by design, because criteria your team never set are criteria your team can't audit. It's less work than it sounds: 5–8 weighted criteria cover most roles well, and there's a free rubric template with weighted tiers and scorecards to start from.
Rubrily is the rubric-first ATS: your rubric, every candidate scored against it, a written justification behind every number. Free for your whole team. Join the waitlist →