Every score on this platform is reproducible from published rules.

The instrument

A participant answers a structured set of questions grouped into dimensions. Most questions use a five-level maturity scale, where each level carries a plain-language description of what it looks like in practice. Others are single or multiple choice, yes/no, numeric or percentage. Some questions are unscored and exist to give reviewers context.

Core and adaptive questions

The instrument has two parts, and only one of them produces the score.

• Core questions are asked of every organization, in the same wording and with the same answer options. They are the entire basis of the score, which is what makes two organizations comparable.
• Adaptive questions are triggered by an organization’s profile — its size, sector, how far its AI work has progressed, whether it builds or buys models. They produce sharper findings and recommendations for that organization, and they never move its headline score in either direction.

Two organizations may therefore answer different numbers of questions and still be measured on the same construct. Which branching rules applied to a submission is recorded with it, so any result can be traced back to the exact question set that produced it.

How a score is produced

• Question score. Each answer maps to a point value between 0 and 100, defined per option in the question bank. Numeric questions are normalised between their configured minimum and maximum.
• Dimension score. Core question scores are combined as a weighted mean within their dimension. A question weighted 1.50 counts 50% more than one weighted 1.00.
• Overall score. Dimension scores are combined as a weighted mean using the dimension weights below, producing a figure from 0 to 100.
• Maturity level. The overall score is mapped onto the bands published further down this page.

Not applicable is not zero

Where a question genuinely does not apply to an organization, the participant may mark it “not applicable”. Such answers are then excluded from that dimension’s denominator entirely, and the weight is redistributed across the remaining questions. Marking something not applicable therefore never penalises an organization for a question it could not reasonably answer.

“Not sure” is recorded, not guessed

A participant who does not know an answer can mark a question “not sure — ask a colleague” instead of guessing. Such a question is never scored and never counted as a zero; it stays on the outstanding list and must be resolved before the assessment can be submitted. Coverage — how much of the core set produced a usable answer — is reported alongside every result, separately from the score itself. A score is a measurement; coverage tells the reader how much weight that measurement can carry.

Completion thresholds

Every required question must be answered before an assessment can be submitted. In addition, a dimension must be at least 60% answered to be scored at all; below that it is reported as “not enough answers” rather than given a low score it has not earned. Dimensions that cannot be scored are excluded from the overall figure rather than counted as zero.

Versioning

Assessment content and scoring configuration are versioned separately, and every submission records both. If the question bank or the scoring rules change afterwards, results already issued are untouched — they remain tied to the version that produced them, and are never silently recalculated. Any deliberate recalculation is recorded in the audit trail.

What this score is and is not

The index is a structured self-assessment of organizational readiness, scored by published rules. It is not a certification, an audit, an accreditation, or a statement of compliance with any law or standard. Submitted answers are self-reported unless a reviewer has separately verified the supporting evidence, and a result says what an organization reported about itself at a point in time.

What AI does and does not do

Artificial intelligence is used, optionally, to explain a question in plainer terms and to draft a narrative summary of a completed result. Both are clearly labelled as AI-generated. AI has no role in producing, adjusting or reviewing a score, and the platform functions completely with those features switched off.