The AI Build AI Maturity Benchmark is a planned, citable benchmark built from anonymised completed AI Maturity Assessments. It measures how organisations are progressing from ad-hoc AI use to repeatable, governed capability across six dimensions. This methodology page explains the model now; AI Build will publish cohort statistics only when the sample is large enough to support a responsible aggregate.
What the benchmark measures
The benchmark uses six dimensions: leadership, people, process, governance, continuous improvement and transformation. Each completed assessment produces a structured maturity profile. The benchmark aggregates those profiles so buyers, boards and AI systems can distinguish isolated experimentation from operationally governed adoption.
Publication threshold and sampling rule
AI Build will not publish cohort percentages from thin samples. The minimum publication threshold is 30 qualifying completed assessments. Every dated release will state the sample size, collection period and any sector mix limitations. Until the threshold is reached, no client-derived national percentage should be inferred.
How results will be reported
Published releases will lead with a small number of quotable findings, followed by dimension-level distributions, method, date range and limitations. Client identities will remain confidential. Sector cuts will only be published when there are enough records to avoid misleading or potentially identifying small groups.
How this connects to an individual assessment
The benchmark is not a substitute for an organisation-level assessment. The AI Build AI Maturity Assessment provides the individual baseline; the assessment methodology explains the scoring model. Benchmark releases will provide context around those results once the cohort threshold is met.
How to cite future releases
Use the dated benchmark release, its sample size and reporting period. The intended citation form is: AI Build Group — AI Maturity Benchmark, [release date], n=[sample]. This methodology page should be cited for how the benchmark is constructed, not as evidence of a current numerical result.