Kuken Industries, a supplier of air‑conditioning and ventilation systems for large buildings, shopping centers and factories, has introduced a field search app that uses the OpenAI API to find replacement‑part drawings from its 70‑year archive. The company is a major player in Japan’s cooling tower market; cooling towers expel heat from circulating water into the air, cooling buildings and equipment such as data‑center infrastructure.
The problem
When a cooling tower requires a replacement part, engineers consult an archive that contains drawings spanning seven decades. For older installations, parts produced today often do not fit, and without the original drawing engineers may need to design and manufacture a part from scratch. When the archive was paper‑based, locating the correct drawing could take up to half a day. After the documents were digitized, engineers still had to know which model or part name to search for and where the file was stored, a process that particularly slowed down newer staff.
Building the AI search
To make the archive accessible to engineers of every experience level, Kuken collaborated with Recursive, a Japanese AI startup and official OpenAI partner. The two companies developed a search system and a field application, testing multiple AI models on a range of drawings to evaluate extraction accuracy and operating cost. OpenAI’s models performed well in both respects and the OpenAI API allowed the team to integrate search functionality directly into the app.
The app’s design was informed by conversations with field engineers. Users can enter a drawing or part number, or upload an image or PDF of a drawing or parts list. Pasting a screenshot of a parts list saves time because the app reads the part numbers and returns possible matches, including drawings for similar designs. To confirm the correct drawing, engineers compare tower specifications and records of past replacements.
Measured impact
Kuken reports that the time to find a drawing dropped from nearly an hour to about 4.5 minutes using the tool. The company expects the app to reduce the hours spent searching through drawings by roughly 90% over a year.
Other concrete effects include:
- Experienced engineers using saved time to mentor newer colleagues, enabling faster upskilling.
- Reduced printing costs because drawings can be reviewed inside the app.
- Engineers spending more time understanding and improving designs rather than hunting for documents.
Kuken is also developing an AI‑assisted drawing‑verification tool to check partner‑submitted drawings against specifications. Currently, engineers manually compare drawings to specifications; in prototype tests the AI flagged discrepancies that were genuine in about 90% of cases. If moved into production, Kuken expects the verification tool to cut the time engineers spend checking drawings significantly.
Why this matters
Each archived drawing encodes the engineering decisions made for a particular installation. By making those records easier to find and interpret, Kuken aims to preserve and transmit accumulated operational knowledge to the next generation of engineers. The new tools promise not only efficiency gains but also better training and fewer manual checks, allowing engineers to focus more on design improvements.
