Evidence · charts
The collection in five charts.
What the 522 examples cover: the type of work, the assets, the year they were published, how well each is evidenced and the time savings they claim.
The charts describe this collection, which reflects what organisations choose to publish. They are not a measure of how widely AI is used or how often it succeeds.
Evidence strength
Most examples describe real work, but few are strongly evidenced.
Of the 522 examples, 28 are Strong, 416 Moderate and 78 Signal. Strong means the main finding is well supported. Moderate means the work is real but important details are missing. Signal means an announcement, plan or claim with little evidence behind it yet. A Strong case can record a failure as well as a success.
Explore Strong cases → · Browse the full collection →
Open the full-size chart →How to read this chart
Each example appears once. These are 38thMove’s assessments of what each source can support. They are not a success rate, and they do not certify every number in a source.
Where AI appears in the work
Inspecting and running assets account for more than a third of examples. Most use predictive AI.
Inspection and condition assessment (87 cases) and asset operation and maintenance (101) lead the collection. About 7 in 10 cases use predictive AI, which spots, forecasts and optimises (for example image recognition, leak detection and scheduling); about 1 in 5 use generative AI, which writes and creates.
Open the full-size chart →How to read this chart
Each of the 522 examples is counted once, under its main type of work. Hatched bars show where 38thMove judged the type of AI because the source did not say. The mix reflects what organisations publish, not how common each use is in the market.
The separate Strong column counts Strong cases already included in each full bar. It is not an additional cohort.
Which assets and stages are covered
Operation and maintenance is the largest group in this collection.
Operation and maintenance accounts for 201 cases, against 69 in design. Buildings and property (152), transport (117) and water (83) are the largest asset groups. Find the cells closest to your own work.
Open the full-size chart →How to read this chart
Each example is counted once, by its main asset type and lifecycle stage where these are known. “Not assigned” covers company-wide and cross-sector work.
Each cell shows the full case count followed by the included Strong count. Darker shading means more cases, not stronger or more positive evidence.
How the collection has grown
Most examples in this collection were first documented from 2024 onwards.
The collection has 75 examples first documented in 2024, 133 in 2025 and 176 so far in 2026. Generative AI accounts for 16 of the 75 examples in 2024 and 57 of the 176 in 2026 (32%).
Open the full-size chart →How to read this chart
Each case is placed in the year of its earliest identified public account; 24 first-observed and 3 public-by references are left out; 40 publication dates are labelled estimates. This is the publication history of a curated collection, not an adoption rate. The type of AI is today’s description; some components may have been added later. 2026 is a part year.
Annual labels identify the Strong cases within each full total. All 28 Strong cases have publication-dated references; no Strong case is among the 27 first-observed/public-by exclusions.
Claimed time savings
Claimed savings on single tasks are often half the time or more. Few sources measure the time spent checking the output.
There are 47 time-saving claims in 40 examples. Most say a task takes half the time or less. Asset operations report other benefits, such as lower energy use, less leakage and fewer call-outs. These are not time measures, so they are not on this chart.
These are the sources’ claims, not results you should expect. Few measure the time spent checking and correcting the output, and tasks and starting points differ, so no average is calculated. Some claims cover overlapping parts of the same process and should not be added together.
Open the full-size chart →How to read this chart
Each point is one claim about one task. Filled points are reported results, hollow points are estimates and arrows are open-ended claims such as “more than 75%”. Connected points show a range the source gave. Several claims can come from one case.
Purple rings identify the two plotted claims from Strong cases. A Strong rating does not independently validate every figure. Select a point in the full-size chart to open its case.
Next
From trial to routine use.
Twelve AI programmes, from trials to wider use, stops and missing follow-up.