Methodology

What we collect

Only publicly available reports of enterprise AI rollouts: vendor customer stories, company press releases, news coverage and government evaluations. For each we record organization, industry, tool, seats or population with access, the usage figure exactly as reported, other metrics (often time saved), results and the source link. Fields the source does not report are left blank.

How we tag figures

  • Metric type: weekly active, daily active, monthly active, "adoption" (undefined or cumulative, such as ever-used or activated), or other (trial or intent figures).
  • Source type: vendor (published by the AI vendor or its partner), company (the deploying organization's own release), independent (press or government reporting).
  • Source pending review: the page blocked automated checks and its content was confirmed only through a search index. These stay visible but are flagged as provisional.

Where a headline figure comes from a study cohort, not the whole organization, we say so. For example, the Accenture GitHub Copilot figures describe a developer study group, not firm-wide usage.

Benchmark calculation

The tool converts each reported figure to a single number (using the stated value, or the lower bound for "over X%"). It then computes the share of comparable figures at or below yours, separately for vendor-published and other sources. It first matches on metric, then on tool. If fewer than three matching points exist, it widens to all figures for that metric and then to all figures, and labels the result accordingly.

Known biases

  • Survivorship and selection: vendors publish their best customers, which is why vendor figures cluster at 80–100%.
  • No common denominator: licensees vs all employees vs pilot group, with the time window rarely stated.
  • Self-reported outcomes: time saved is almost always survey-based.
  • Coverage skew: banking, telecom and Japanese firms are over-represented. Healthcare, retail and the public sector are thin.
  • Seat counts are often announcements, not confirmed active seats.

Independence

Sponsors never influence data, tags or rankings. Sponsored case-study listings are always labeled and tagged as vendor-sourced. Corrections are welcome at [email protected].