Competency-based assessment judges whether a person can actually perform a task to a defined standard - competent or not yet competent - rather than scoring them on a percentage, ranking them against peers, or rewarding time spent studying. The unit of currency is demonstrated capability: can they do the thing, to the required level, with evidence to prove it. It is built on a binary, criterion-referenced decision against explicit performance criteria, which is why it dominates vocational training, professional licensing, and any setting where the result is a claim that someone is safe and able to do real work.
The core idea: can they do it, yes or no
Traditional grading asks "how well did they score?" and produces a number on a scale. Competency-based assessment asks "can they perform to the standard?" and produces a decision. There is no 72%. There is "competent" or "not yet competent" - and the phrase "not yet" matters, because the model assumes the learner can get there with more practice and another attempt, not that they have failed permanently.
This reframes what learning is for. Time in a seat is not the goal; demonstrated competence is. Two learners might reach competence at very different speeds, and that is fine - the standard is fixed, the path and pace are not. It is the opposite of a system where everyone moves at the same rate and is sorted by who scored highest.
What competency-based assessment requires
- Explicit performance criteria. The standard has to be defined and observable before anyone is assessed - what competent performance looks like, spelled out. Vague criteria make a binary decision arbitrary.
- Evidence of performance. The decision rests on the learner actually demonstrating the skill - a real task, a work sample, a project, an observed performance - not on a proxy like recalling facts about the skill.
- Criterion-referenced judgement. Each learner is judged against the standard, never against other learners. No curve, no quota.
- Defensible records. Because the outcome is consequential, the evidence and the reasoning behind each decision have to be documented and auditable.
The rules of evidence apply throughout: evidence must be valid, sufficient, authentic, and current. A competency decision is only as strong as the evidence and the record behind it.
Where it is used
Vocational education and training is the clearest home - in Australia, learners are assessed as competent or not yet competent against units of competency in nationally defined training packages. But the model is everywhere competence has to be guaranteed: trade licensing, healthcare certification, aviation, professional accreditation, and increasingly corporate capability frameworks and bootcamps that certify someone is job-ready. Anywhere the question is "can this person safely do the job?", competency-based assessment is the natural fit.
The challenge, and where AI marking helps
Competency-based assessment is evidence-heavy and judgement-heavy by design, which makes it labour-intensive. Demonstrating competence usually means open-ended work - a real task, a written response, a portfolio, a recorded performance - and a qualified assessor has to read all of it and judge it against the criteria. That is exactly the work that does not scale by adding more multiple-choice questions.
AI rubric marking fits this model well because competency criteria are essentially a rubric: defined levels of performance with evidence requirements. The model reads the learner's actual demonstration, maps it against each criterion, and cites the specific evidence for a competent or not-yet-competent call - then a qualified assessor reviews and signs off. The decision stays human and accountable, as it must for a consequential outcome, while the grind of reading and mapping every submission is handled. That keeps the assessment defensible and the assessor focused on genuine judgement calls.
This is the model Scorafy is built for - open-ended evidence marked against your competency criteria, with cited support and human sign-off. For VET and RTO teams it maps directly onto units of competency; see assessment for VET trainers and tools for RTOs. To define the criteria well, start with how to write an assessment rubric, or try it on a real submission.