Raeburn Institute methodology

UK Enterprise AI Governance Benchmark methodology

The benchmark is designed as an open, repeatable longitudinal measure of practical governance maturity—not a certification scheme and not a claim that voluntary respondents perfectly represent every UK organisation.

Scoring

  • 20 core statements across 10 equally weighted governance dimensions.
  • Each statement is scored 0–4: Not in place, Ad hoc, Partly defined, Defined, Evidenced & operating.
  • Overall score is the achieved points divided by the maximum possible points, expressed from 0–100.
  • Annual reporting will distinguish self-reported governance maturity from independently verified controls.
  • No percentile or UK-wide representativeness claim is made until the dataset supports it.
  • Cohort cuts may include broad organisation-size, sector and AI-adoption bands only when sample sizes are sufficient to reduce disclosure risk.

Anonymous research collection

The live research form does not ask for organisation name, respondent name, email address, phone number or free-text organisational descriptions. Optional cohort fields use broad categories only. Benchmark responses are stored separately from commercial lead data, and no raw organisation-level response dataset is published.

Minimum publication safeguards

The annual report must state the number of eligible responses, fieldwork dates, recruitment channels, completion/exclusion rules, missing-data handling and limitations. Small cohort results should be suppressed. Findings must be labelled voluntary-sample findings unless a defensible sampling design supports stronger inference.

Annual outputs

Planned outputs include overall maturity distribution, dimension-level maturity, control adoption rates, year-on-year movement where comparable, broad cohort comparisons where safe, and a transparent limitations section. The first annual report will only be released once there is enough eligible data for useful aggregate analysis.