About

A shared, independent reading of business AI research.

AI Unite reads the reports that consultancies, universities, companies, government and investors publish about AI in business, and writes short, sourced summaries so leaders and practitioners can compare them side by side.

How reports are chosen

A report is added to the library when it meets all of these criteria.

  1. 01It is published by a named organization, not an anonymous source.
  2. 02It is about AI in business: strategy, adoption, workforce, value or governance.
  3. 03It describes its method, such as who was surveyed, how many and when.
  4. 04The full report is publicly available, free or paid.

How summaries are written

Each summary follows the same steps, so reports can be compared on equal terms.

  1. 01We read the full report, not only the press release.
  2. 02We record the key findings and, where possible, the page each one comes from.
  3. 03We note the method, the sample, any limitations, and what the publisher sells.
  4. 04Every number is checked against the publisher’s own page or PDF before it is published. Charts are redrawn from published figures, never copied.

How we rate trust

Each report gets a trust rating of high, medium or low. It reflects how much weight we think its headline numbers can bear, not whether we agree with them.

  1. 01Sample: how many were asked, and whether they resemble the companies the report talks about.
  2. 02Method: whether the report says who was asked, how, and when.
  3. 03Independence: whether the publisher sells AI products or AI services. That does not make data wrong, but it shapes which questions get asked and which number leads.

Stages of adoption

AI adoption surveys disagree mostly because they count different things. We tag each headline number with the stage it measures, so figures can be compared fairly.

  1. 01Tried it: any use of ai.
  2. 02Approved: sanctioned in a function.
  3. 03In core work: running in key processes.
  4. 04Pays off: measurable profit impact.

Independence

AI Unite does not accept payment from publishers to include, feature or rank a report. Summaries describe what a report found and how; they do not endorse its conclusions. Every summary links to the original so readers can check it.

Corrections

If a summary misstates a finding, email corrections@aiunite.org with the report and page number. Corrected summaries show the date they were updated.

Open data and reuse

Our summaries are licensed CC BY 4.0: you may quote and reuse them with a link back. The full library is available as structured data. The reports themselves belong to their publishers; please cite and link to the originals.

Contribute to the library

Suggest a report we’re missing, or tell us about a correction.

Suggest a report