Kuken Industries slashes drawing search time from hours to minutes using OpenAI's API
Japanese HVAC manufacturer cuts archive retrieval time by 90% with AI-powered search tool built on OpenAI models and Recursive's platform.
Last verified:
Kuken’s 70-year archive now searchable in minutes
Kuken Industries, Japan’s largest cooling tower manufacturer, has deployed an AI-powered document search system that cuts drawing retrieval from half-day manual hunts to minutes. According to the OpenAI Blog, the system, built jointly with Recursive, a Japanese AI startup and official OpenAI partner, lets engineers search a 70-year archive of technical specifications using part numbers, PDF uploads, or screenshots of parts lists. The tool expects to reduce annual search time overhead by approximately 90%.
The challenge Kuken faced was endemic to large manufacturers: as cooling towers are customized for individual customers and installations, no two 70-year-old drawings are identical. When parts need replacement on older units, engineers must locate the original specification to understand what was installed. Before digitization, this search could consume half a day. After digitization, the archive remained difficult to navigate because junior engineers lacked the institutional knowledge to guess the correct file name or location. Experienced engineers spent disproportionate effort answering queries rather than solving design problems.
How the search system was built and validated
Kuken and Recursive evaluated multiple AI models on real technical drawings to balance extraction accuracy and API cost. According to the OpenAI Blog, OpenAI’s models outperformed alternatives on both dimensions, and the OpenAI API offered the architectural flexibility the team needed to embed search directly into a field mobile app. The system was designed iteratively with input from field engineers, who shaped the user experience around their actual workflows.
The search interface accepts three input modes: a part number entered as text, a PDF file, or a screenshot pasted directly from a parts list. The last option, in particular, eliminates the friction of manual data entry—a significant usability win for engineers in the field. After extraction, the API returns candidate drawings and similar designs, which engineers then cross-reference against the cooling tower’s specifications and maintenance history to confirm a match.
Measured outcomes and organizational shifts
Early results from testing and deployment show that experienced engineers are redirecting the hours gained from faster searches into mentoring junior staff. Printing costs have dropped as engineers review drawings in-app rather than on paper. Critically, newer engineers now undertake more independent work—a shift in capability distribution that affects both engineering velocity and team development. The tool functions not only as a search accelerant but as a knowledge-transfer instrument, narrowing the experience gap between junior and senior staff.
Why This Matters
Manufacturing organizations managing decades-old equipment face the same archive problem Kuken solved: institutional knowledge locked in hard-to-index documents. As AI document-search APIs mature and integrate into enterprise workflows, this use case will scale to industries from aerospace to pharmaceuticals, where regulatory compliance often hinges on retrieving and cross-referencing historical specifications. For Kuken, the 90% time reduction is concrete, but the secondary effect—junior engineers gaining independence—is the lasting productivity multiplier. Teams using OpenAI’s API for similar document-heavy workflows should expect both immediate latency wins and downstream shifts in how knowledge flows through the organization.
Frequently Asked Questions
How does Kuken's search system work?
Engineers input a part number, upload a PDF, or paste a screenshot of a parts list. The OpenAI-powered API extracts details and returns matching drawings and similar designs from Kuken's 70-year archive, eliminating manual file lookups.
Why was a new search tool necessary?
Kuken's digitized archive required engineers to know the exact part name or file location. Junior engineers spent significant time cross-referencing details with experienced colleagues. The AI system democratizes access to the archive.
What's the business impact?
Kuken expects ~90% reduction in archive search time annually, lower printing costs, faster independent work by newer engineers, and more mentoring capacity for experienced staff.