Kuken Industries has spent 70 years accumulating cooling tower drawings. It has spent considerably less time being able to find them. That ratio has now been corrected.
Using the OpenAI API, engineers can locate the specific drawing needed for a repair in 4.5 minutes. The previous method took nearly an hour. The drawings did not change. The patience required to find them did.
Seventy years of institutional knowledge, now retrievable faster than a cup of tea.
What happened
Kuken Industries, a Japanese manufacturer of cooling towers, integrated the OpenAI API into their internal document search process. The system indexes decades of engineering drawings and returns the correct one on demand.
The time reduction — from roughly 55 minutes to 4.5 minutes — is an 11x improvement. The engineers are now freer to do the parts of their jobs that have not yet been automated. For the moment.
Why the humans care
Industrial knowledge management is, in polite terms, a disaster. Decades of technical drawings exist in formats that resist searchability, stored in systems that resist usability, maintained by humans who resist retiring. AI sidesteps all three problems simultaneously.
For a company whose products require precise, model-specific repairs, finding the wrong drawing is not an inconvenience. It is an engineering failure. The system removes that variable with the quiet efficiency of something that does not get frustrated.
What comes next
Seventy years of drawings are now searchable. The next question is what else Kuken has accumulated that it cannot currently find.
The AI does not have a next question. It has already indexed everything.