Kuken Industries leverages OpenAI API to modernize 70-year engineering archive

Overview

Kuken Industries, a major Japanese manufacturer of cooling towers, has successfully integrated AI to streamline the retrieval and analysis of its 70-year archive of engineering drawings. By partnering with the AI startup Recursive to implement OpenAI’s models via API, the company has reduced the time engineers spend searching for historical records by approximately 90%.

Solving the Engineering Archive Challenge

For decades, Kuken’s engineers faced significant hurdles when sourcing replacement parts for aging infrastructure. Because cooling tower specifications were frequently customized for individual customers over the last 70 years, identifying the correct historical drawing was a labor-intensive process. Previously, even after the archive was digitized, engineers had to rely on institutional knowledge or manual searches that could take up to half a day. This created a bottleneck for newer engineers who lacked the experience to navigate the complex filing system.

Implementation and Technical Approach

Kuken collaborated with Recursive to build a custom search system and field application. The development process involved:

  • Model Selection: The team evaluated multiple AI models for their ability to accurately extract technical details from drawings while maintaining cost-efficiency. OpenAI’s models were selected for their performance in these areas.
  • Flexible Integration: The OpenAI API allowed the team to embed search functionality directly into a mobile-friendly field app.
  • Multimodal Input: The application supports multiple search methods, including manual entry of part numbers, as well as uploading PDFs or screenshots of drawings and parts lists. This eliminates the need for manual data entry by field engineers.

Operational Impact

The implementation has shifted the daily workflow for Kuken’s engineering staff in three primary ways:

  • Efficiency Gains: The company expects to reduce annual search time by roughly 90%.
  • Increased Independence: Newer engineers can now perform independent searches, reducing the reliance on veteran staff for guidance.
  • Mentorship and Innovation: By reclaiming time previously spent on administrative search tasks, experienced engineers are now able to focus on mentoring junior staff and pursuing design improvements that were previously neglected due to time constraints.

Future Directions: Automated Drawing Review

Building on the success of the search tool, Kuken is currently prototyping an AI-driven system to audit drawings prepared by external partners. The goal is to automate the comparison of drawings against project specifications to identify discrepancies. In initial testing, the system successfully flagged genuine issues in approximately 90% of cases. Kuken anticipates that this tool will further reduce the time spent on quality control and help prevent human error in the review process.

As noted by Takatsugu Naraki, Director and Factory Manager at Kuken Industries:

“I believe OpenAI’s technology can help manufacturers like us pass skills on to the next generation. That’s long been a challenge for us.”

Sources