Why this section exists
Much advice on AI in business rests on announcements and opinion. This site separates what has been tested from what has only been said. This section shows how.
- How we rate evidence The four evidence levels, the two flags, what does not count as evidence, and how often ratings are reviewed. Version 1.1.
- Studies The studies behind the practice ratings: how each was done, what it found, and where the result stops applying.
- Company cases Named companies with numbers and a public source: what they did and what happened next. Failures are included.
- Claim checks Widely repeated figures and sources, traced back to the primary source. Each has a verdict and a sentence you can safely cite.
- Evidence gap 58% of the AI success stories in the case set rest only on the company’s own figures. How the figure is computed, its limits and how to cite it.
Looking for questionnaires and tests to use in your own company? See assessments. To see a method applied step by step, see the worked examples.