AI-graded assessment is a hiring-evaluation method where a candidate's answer to a work-sample question is scored by AI against a rubric the hiring team defined in advance — criterion by criterion, with the evidence for each score shown — and a person on the hiring team confirms the result before it counts toward a decision. It's narrower than "AI hiring" in general: the AI's job is to grade against your standard, not to decide who advances.
The three parts that make it "AI-graded," not "AI-decided"
The term gets used loosely, but a real AI-graded assessment has three specific parts, in this order. Skip any one of them and you have something else — a black-box score, or an unstructured AI opinion.
- A rubric written and frozen before anyone answers — the criteria, not a grader's after-the-fact impression, define what a strong response looks like.
- AI grading that works criterion by criterion, marking each as met, partial, or not shown, with the evidence it drew from — not one opaque number.
- A human confirmation step where a person on the hiring team reviews that reasoning and confirms the result. The AI's grade is a recommendation; it is never the final word.
What it looks like end to end
Before any candidate answers, the hiring team writes the rubric for the question — the specific criteria that matter for the role — and it's locked in for that hiring run. When a candidate submits an answer, the AI grades it against that fixed rubric, criterion by criterion, and writes a short explanation of why each was scored the way it was. A member of the hiring team then reviews that reasoning and confirms the grade before it feeds into any decision. Every step — the grade, the reasoning, the confirmation — is written to an immutable audit trail, so the process can be explained after the fact, not just trusted at the time.
AI grades, you decide
That line isn't a tagline — it's the actual division of labor. AI produces a consistent, evidenced grade against your criteria. A person decides what that grade means for the candidate. On Gradogram, this same flow runs under all four live assessment types: Knowledge, System Design, Behavioral, and Culture Fit.
What it's deliberately not
AI-graded assessment is not an AI that decides who gets hired — the confirmation step exists specifically so a consequential call about a candidate always has a person behind it. It's also not a single opaque score with no way to check it: if a platform can't show you why a candidate got the grade they did, criterion by criterion, it isn't doing this. And it isn't a claim that the AI is infallible — anything it can't confidently grade should be flagged for manual review rather than scored anyway.
Why hiring teams are moving to it
The pain it solves is consistency at volume without losing the ability to explain a decision. A rubric applied by hand across fifty candidates drifts — a grader's standard shifts by candidate thirty. A rubric applied by AI against a frozen standard doesn't drift, and because every grade comes with its reasoning attached, a hiring manager (or, if it ever comes to it, a regulator) can see exactly why a given candidate scored the way they did. That combination — consistent grading, visible reasoning, a human decision — is the actual value, not "AI hiring" as a category on its own.
If you want the deeper mechanics — how the rubric is structured, what the AI's confidence read means, how overrides get logged — the full walkthrough is in how rubric-based AI grading works. This post is the short answer; that one is the whole path from a candidate's answer to a confirmed decision. And if you're the one writing the rubric in the first place, here's how to write one that's actually gradable.