OpenAI case study: Kuken searches 70 years of cooling tower drawings in 4.5 minutes

OpenAI published a customer story on November 9 describing how cooling tower manufacturer Kuken Industries uses the OpenAI API to search 70 years of cooling tower drawings.

The headline number is simple: engineers can now find the drawing they need for a repair in 4.5 minutes, down from nearly an hour.

Cooling towers are long-lived equipment, and a repair, retrofit or spare-part replacement can send engineers back to design drawings made long before modern document systems existed. Those archives are large, the numbering schemes are old, and finding the right sheet is one of the slowest parts of the job.

Kuken routes that archive through the OpenAI API, putting the model in charge of the retrieval step. The case does not describe the technical implementation, the scale of usage or the deployment architecture.

The value of a case like this is not the model itself but the fact that it sits inside a workflow with a measurable hourly cost: near-hour retrieval becomes 4.5 minutes, and the saving converts directly into downtime and labor.

For manufacturers and engineering firms, drawings, work orders and maintenance logs are classic long-tail knowledge assets — messy formats, decades of history and dense domain vocabulary, exactly where generic search tools struggle.

What the case does not disclose matters too: retrieval accuracy, access controls and whether an engineer still reviews each result. In maintenance work, the wrong drawing is worse than no drawing.

The next question is whether this kind of archive search scales from a single customer story into a standard product, and whether vendors publish more on accuracy, cost and data governance.