A named author
Every finding is attributable to a person who can be asked to defend it. Nothing is issued as the impersonal output of a system.
Method
Human-supervised from the first question to the last caveat.
NSAG has no product that runs without a person in it. That is not a limitation being apologised for — it is the finding the framework is built on. A governance question that can be answered by a form was not a governance question. This page describes how the work is actually done, so you can judge it before contacting anyone.
What is not on offer. There is no instant tier report, no automated analyst, no guaranteed benchmark, and no self-serve paid engagement. The fifteen module self-assessments are retired and cannot be taken. Nothing on this site enrols you, charges you, or produces a deliverable without a person agreeing to produce it first.
The unit of work is a decision, not a topic
Most requests arrive as a subject — “burnout”, “our AI tool”, “the new building”. A subject cannot be answered; it can only be discussed. The first thing any engagement does is convert the subject into a decision somebody actually has to make, with a date attached and a person who owns it.
This is the step most often skipped, and skipping it is why governance reviews so frequently end in a document nobody uses. If there is no decision, the honest answer is that an advisory engagement is the wrong instrument, and you will be told so rather than sold something.
What the framework brings to it
Every module in the library states the same three things: the evidence base for one institutional design problem, six governance dimensions that problem turns on, and observable criteria for each of four tiers. That structure is what makes a review reproducible rather than a matter of taste.
The tiers describe governance infrastructure, not intentions. The test at each level is a documentary one:
| Tier | The question it has to survive |
|---|---|
| Pioneering | Can the institution produce written evidence of implementation, measurement, and leadership accountability — today, without advance notice? |
| Emerging | Has the institution documented at least one governance change in this domain and measured its effect? |
| Developing | Would the governance change survive the departure of the person currently driving it? |
| Early Stage | Does the institution know what it is currently producing in this domain — in measurable, documentable terms? |
Most institutions are Early Stage in most domains. That is a structural description, not a judgement, and saying so plainly is part of the method: a review that flatters everyone measures nothing.
How an engagement actually runs
-
Scope
The decision is named, along with who owns it, which jurisdiction governs it, who is affected, and what is explicitly out of scope. Exclusions are written down at the start, because an unstated exclusion later reads as an oversight.
-
Evidence boundary
Before anything is read, the boundary is fixed: which literature, which internal documents, which time window, and what will not be looked at. A finding is only as good as the boundary it was produced inside, and a boundary declared afterwards is indistinguishable from a boundary drawn to fit the conclusion.
-
Human review
A person reads the material and writes down assumptions and uncertainty as they go — not at the end. Where the evidence is thin, the review says the evidence is thin. Where two sources disagree, both are recorded rather than reconciled by preference.
-
Deliverable
Format, length, timeline, price, and follow-up are agreed before work begins, not inferred from a menu. What is delivered states its own evidence, its date, who authored it, and its limitations, so that a reader six months later can tell what it did and did not cover.
What human supervision means concretely
Declared limits
The limitations section is written before the recommendations, not appended to them. If it is uncomfortable to write, that is the signal it matters.
Dated evidence
Sources carry the date they were consulted. Governance evidence rots — regulations change, vendors change, staff leave — and an undated claim hides that.
A stated boundary
What was excluded is published alongside what was examined, so a reader can judge the shape of the gap rather than assume there is none.
What NSAG declines
Refusals are part of the method, and they are consistent:
- Certification, accreditation, or a score an institution can display. The framework is not a validated psychometric instrument and does not accredit anyone. A badge would misrepresent what the evidence supports.
- Clinical or legal advice. The modules describe institutional design, not individual care and not individual legal positions.
- A benchmark across institutions. Comparing institutions requires an approved cohort methodology, consent, minimum-cell rules, and a de-identification review. None of those exist yet, so no benchmark is offered.
- A recommendation produced without reading the institution's own documents. The tier tests are documentary by design; an assessment that skips the documents is a survey of impressions.
Where to go from here
The material is public and needs no conversation. Read the framework and tier definitions, the fifteen module scopes, or the evidence and resource pages. The operations matrix states, module by module, what is public, what is retired, and what is delivered by a person.
If there is a decision with a date on it, a scope conversation is the right next step. If there is not, reading is genuinely the better use of your time.
Discuss a scope What an engagement delivers
This page describes NSAG's own delivery method and states no third-party factual claims. Tier tests quoted verbatim from the published NSAG framework. Reviewed 29 August 2026.