Skip to main content

Quick Overview

This case study video produced by OpenAI highlights a collaboration between Japanese cooling equipment manufacturer Kuken Industries and AI development firm Recursive. Factory managers and engineers describe their operational challenges with legacy archives and explain why they implemented AI-assisted search tools on factory documentation.

Key Points

  • 1.Kuken Industries partnered with Recursive to implement an OpenAI-powered search system across seventy years of cooling equipment engineering drawings.
  • 2.Digitizing paper records had previously failed to solve retrieval bottlenecks because workers still relied on senior staff to navigate unstructured archives.
  • 3.Recursive selected OpenAI models after internal benchmark testing showed superior performance when querying unstructured technical records.
  • 4.Drawing searches that previously required up to half a day or sixty minutes were cut to approximately ten minutes.
  • 5.Time saved from manual document searches is redirected toward verifying drawing accuracy, tackling complex projects, and mentoring junior employees.

Summary

Kuken Industries has manufactured air conditioning and cooling equipment for seventy years, supplying specialized machinery such as cooling towers to large facilities including shopping centers and museums. Over seven decades of production, the internal engineering and structural designs of these cooling towers have continuously evolved. Consequently, systems manufactured ten or twenty years ago differ substantially from modern units. When these older installations require service, maintenance technicians cannot simply assume standard modern dimensions or parts; they must locate the exact historical blueprints and assembly records to understand how each specific machine was built.

Historically, identifying and retrieving the correct documentation created a significant operational bottleneck. Although Kuken Industries had begun digitizing its paper archives, simple image scanning did not establish a uniform search system. Staff members could not reliably query the digitized files without specialized institutional familiarity. Instead, maintenance teams relied heavily on the accumulated experience and judgment of senior personnel to guess which binders, files, or drawing sets corresponded to a particular machine. Depending on the age of the equipment and the volume of archived paperwork, locating a single drawing could take up to half a day, while routine lookups routinely required sixty minutes. Because younger technicians lacked the long-term context of veteran workers, passing down operational know-how remained a persistent challenge across the factory.

To overcome this barrier, Kuken Industries collaborated with Recursive, an OpenAI partner firm, to develop a specialized search system for engineering drawings. Recursive's solution connected Kuken's seventy-year archive of technical records with OpenAI models. Recursive selected OpenAI models after running internal benchmark tests, which demonstrated that the models excelled at parsing highly unstructured data and retrieving precise technical details from dense records. The resulting search application allows factory technicians to input a specific part number shown on an assembly diagram. The system then searches across seven decades of documentation to identify every engineering drawing and record containing that specific part number.

The deployment of the AI tool drastically reduced retrieval times across the maintenance division. Lookups that previously took sixty minutes were cut to approximately ten minutes, eliminating lengthy manual investigations that previously occupied half a day. Maintenance technicians noted that reducing file lookup times gave them substantially more capacity to inspect blueprints, verify technical accuracy, and evaluate engineering improvements. Senior staff repurposed the saved hours to mentor younger workers and tackle more demanding, high-level client projects. Factory leadership emphasized that the system adheres to the core philosophy that human workers must retain final decision-making authority, using artificial intelligence to handle information retrieval and preserve decades of manufacturing knowledge for the next generation.

Preserving Manufacturing Expertise and Legacy Drawings

Kuken Industries has manufactured cooling equipment and cooling towers for seventy years, resulting in a large variety of historical configurations. Maintenance on older installations requires reviewing original engineering drawings, which historically depended on the memory and experience of veteran workers, creating a bottleneck in transferring knowledge to junior staff.

Building an AI Search Tool with Recursive and OpenAI

Kuken Industries collaborated with AI consultancy Recursive to connect seventy years of digitized engineering documents to OpenAI models. Recursive evaluated internal benchmarks to choose OpenAI models for their ability to retrieve precise part numbers and technical details from highly unstructured documentation.

Reducing Search Times and Supporting Human Decisions

The new retrieval system reduced drawing searches from up to half a day or sixty minutes down to roughly ten minutes. Factory leadership emphasized that AI supports rather than replaces human staff, freeing technicians to mentor younger colleagues, verify technical accuracy, and make informed operational decisions.

The Bottom Line

The video demonstrates how Kuken Industries and Recursive applied OpenAI models to turn decades of unstructured manufacturing drawings into an easily searchable digital archive. It shows that automating component retrieval significantly shortens search times while allowing experienced technicians to transfer institutional knowledge to newer personnel. The presentation lands on the principle that AI should support rather than replace human decision-making in technical workflows, while leaving open how the company plans to expand the system beyond part number lookups.

FAQ

What is Kuken Industries and what cooling equipment does Kuken Industries manufacture?

Kuken Industries is a manufacturing company that has built air conditioning and cooling equipment for seventy years, including large cooling towers used in facilities such as shopping centers and museums.

Why was digitizing paper records insufficient on its own for Kuken Industries workers searching historical drawings?

Simple digitization did not ensure equal searchability or repeatable document discovery, leaving workers reliant on the experience and judgment of senior staff to determine which specific records and drawings to consult.

Why did partner company Recursive choose OpenAI models for the Kuken Industries drawing search system?

Recursive selected OpenAI models because internal benchmark results proved that the models excelled at accurately retrieving relevant technical information from highly unstructured document archives.

How does the engineering drawing search tool developed by Recursive identify documents for Kuken Industries technicians?

Users enter a specific component part number into the system, and the AI searches through seventy years of archived documentation to display every document containing that number.

How much time did document searches take before and after Kuken Industries adopted the AI drawing search system?

Searches that previously required up to half a day or roughly sixty minutes were reduced to approximately ten minutes using the AI retrieval system.

How do Kuken Industries technicians use the time saved by the AI drawing search tool?

Technicians use the reclaimed time to verify drawings, evaluate improvements, handle more demanding engineering projects, and mentor junior colleagues.

Worth watching for

Manufacturing managers, digital transformation leads, and operations engineers seeking practical case studies on applying generative AI to legacy archives and technical know-how transfer.

  • manufacturing
  • openai
  • recursive
  • document-search
  • knowledge-transfer