Assessment rubric

Every assessment SAAII sets is marked against the same four-dimension rubric. It is published here in full because we think you should know exactly how your work will be judged before you submit it — and because the rigour of the rubric is one of the things that makes a SAAII credential worth having.

The four dimensions

What each score means

How each dimension is scored, in detail

Regulatory accuracy (30%)

Assesses whether the substantive law, regulation, and authoritative guidance in your answer is correct, current, and appropriately cited.

• 25–30 (Distinction) — Regulatory content is accurate and current throughout. Sources are cited where a professional reader would want to check them. Nuance and edge cases are recognised. No material misstatement.
• 18–24 (Pass) — Regulatory content is broadly accurate with only minor slips. Key sources are cited. Main points are correct even where nuance is missing.
• 12–17 (Refer) — Material misstatements or omissions in more than one place, or a pattern of citing outdated guidance where current guidance materially differs.
• 0–11 (Fail) — Regulatory content is substantially wrong in ways that would mislead a reader taking action on your advice.

Applied judgement (30%)

Assesses the quality of reasoning applied to the specific situation posed. Not “did you get the right answer” — often there isn’t one — but “did you reach a defensible conclusion by defensible means”.

• 25–30 (Distinction) — Identifies the real question, weighs the material considerations, acknowledges what is uncertain, and reaches a clear conclusion that a senior colleague would recognise as sound.
• 18–24 (Pass) — Reaches a reasonable conclusion with adequate reasoning. May miss a consideration or two but does not miss the central issue.
• 12–17 (Refer) — Misidentifies the central issue, or reaches a conclusion the reasoning does not support.
• 0–11 (Fail) — Reasoning is absent, incoherent, or would produce serious harm if acted upon.

Artefact quality (25%)

Assesses whether the deliverable meets the professional standards of the artefact type (memo, board paper, incident report, regulator response, etc.). SAAII assessments are set as artefacts precisely because that is the work you will actually do.

• 21–25 (Distinction) — Reads as professional practice work. Structure, tone, and formatting are appropriate to the audience and use case. A senior reader would circulate it.
• 15–20 (Pass) — Meets professional standards. Some structural or tonal weaknesses but not so significant as to undermine the artefact’s usefulness.
• 10–14 (Refer) — Artefact would need substantial rework before being usable in practice — misjudged tone, poor structure, wrong format for the audience.
• 0–9 (Fail) — Artefact is not recognisably of the type asked for, or is unfit to be shown to a professional audience.

Communication (15%)

Assesses clarity, structure, and audience-appropriateness. Even the best analysis is worth little if the reader cannot act on it.

• 13–15 (Distinction) — Prose is clear, well-structured, and pitched precisely for its audience. A busy senior reader could act on it after one reading.
• 9–12 (Pass) — Prose is clear enough. Structure supports the argument. Occasional lapses in clarity or pitching but nothing that materially obstructs understanding.
• 6–8 (Refer) — Prose is unclear or misjudged for the audience. Reader has to work harder than they should to follow the argument.
• 0–5 (Fail) — Prose is not fit for professional use.

Who marks your work

Every assessment is marked by a named human examiner drawn from SAAII’s pool of practitioner-graders — professionals with active or recent standing in the field. A sample of every examiner’s marking is reviewed by a moderator to ensure consistency across cohorts. Where a Learner requests a review, a second independent examiner marks the work afresh.

Turnaround

Marks and examiner comments are returned within 15 working days of submission. Review requests are turned around in the same window.

Comments and feedback

Every mark is accompanied by comments organised against the four dimensions above, so you can see exactly where the strengths and weaknesses of your work lie. Comments are designed to be actionable — if you have to resit, you’ll know precisely what to fix.

Why this way

Multiple-choice questions can test whether you have seen material before. They cannot test whether you can produce the kind of work the field actually asks of you. AI governance is a judgement discipline — the question is almost never “which of these four options is correct” and almost always “which combination of considerations should shape what we do here”. A rubric that pays 60% of its marks to accuracy plus judgement, and 40% to the quality of the deliverable and its communication, reflects the actual weighting of the work.

This is also why the rubric is public. If we published only the pass mark and the outcome, you would have to take our word for the quality bar. Publishing the rubric — and, in each Programme’s materials, published sample answers marked against it — lets you see the bar for yourself.