All cases
Search 522 examples of AI at work.
The examples range from planning offices to bridge inspections. Each one gives a short summary: who did what, where and when, and how well it is evidenced.
The full collection goes further: the original sources, my assessment of what each case supports and where it falls short, and examples not published here. Ask me about it.
What does evidence strength mean?
Strong: actual use, a completed test or substantive investigation supports a defensible central finding with an appropriate method. Moderate: concrete work is documented, but important gaps remain in the central result or validation. Signal: mainly intention, promotion or an inadequately established AI application. Positive and negative findings face the same test. Strength does not measure the size of a claimed benefit; independent evaluation is identified separately.
Predictive AI spots, forecasts and optimises, for example image recognition, leak detection and scheduling. Generative AI writes and creates, for example chat assistants and drafting. A few cases use both; for some, the source does not say. A location can be an organisation’s office rather than the site of the work. Want a deeper analysis? Ask the database your question.
Selected positive examples are a curated list across evidence strengths, with the reported finding, provenance and limit shown separately. Other cases have not been classified as negative. All cases and qualified figures remain available.
Showing a hand-picked starting set first, then the newest examples.
This is a curated collection of what has been published, not a count of deployments or proven benefits.
From research to action
Find out what AI is worth to you.
Published cases are a starting point. The 38thMove methods show how to test an idea on your own work and decide whether to invest.