Tech • AI • Robotics • Game
Kuken Industries searches 70 years of drawings in minutes
6/10TL;DR
Kuh Industrial is using OpenAI-based search with partner Recursive to turn decades of engineering records into a faster, more accessible knowledge system aimed at easing skills transfer in manufacturing.
KEY POINTS
A persistent skills-transfer problem
Manufacturing firms often struggle to pass veteran workers’ skills and know-how to younger employees. At Kuh Industrial, the issue is especially important because work on air-conditioning and heat-related machinery depends on understanding design changes accumulated over 10 to 20 years or more.
A long history of complex equipment
Kuh Industrial has spent decades building machinery tied to air conditioning and thermal systems. Its equipment is commonly used to cool large public spaces such as museums and shopping centers, where accurate access to past drawings and technical records is critical.
Digitization alone did not solve retrieval
The company has been converting paper materials into data, but digitization by itself did not make the information easy to use. Finding the right document still depended heavily on experienced employees who knew which records and drawings to check, limiting reproducibility for less experienced staff.
AI linked to 70 years of records
Working with Recursive, an OpenAI partner, the company is building a system that connects 70 years of internal data with AI tools. The initial focus is drawing retrieval, allowing users to enter a specific part number and identify which documents across decades contain that number.
Why OpenAI was chosen
OpenAI models were adopted after internal benchmarking showed strong results. The main advantage was performance on highly unstructured data, where the system was able to retrieve relevant information more effectively than previous approaches.
Search time cut sharply
Engineers said locating a drawing could previously take 60 minutes and, in some cases, as much as half a day depending on the age and volume of records. With the new search function, that process has been reduced to about 10 minutes in some cases.
More time for judgment and training
The time savings are being redirected toward higher-value work. Instead of spending hours hunting for documents, staff can spend more time judging whether a drawing is correct, deciding how it should be improved, training younger employees, and handling more advanced projects.
Humans retain final authority
The company says the system is designed to support, not replace, human decision-making. Final judgments remain with employees, while AI is positioned as a tool to expand the time available for analysis and decisions rather than automate them away.
CONCLUSION
The project shows how AI can address a core manufacturing weakness by making institutional knowledge easier to find and use. If the approach scales, it could help companies preserve expertise while improving productivity and workforce development.
Ask a question