Kuken Industries Cuts Drawing Search Time to 4.5 Minutes with AI

Kuken Industries, a leading manufacturer in Japan's cooling tower market, has implemented an AI-powered search system to streamline its operations. By leveraging the OpenAI API, the company has reduced the time required to locate technical drawings for replacement parts from nearly an hour to just 4.5 minutes. This innovation is projected to cut annual man-hours spent on such searches by 90%, significantly enhancing operational efficiency.

Challenges in Accessing Decades of Technical Drawings

With over 70 years of archived cooling tower drawings, Kuken Industries faced challenges in locating specific records, particularly for newer engineers. Historically, finding the correct drawing often required consulting experienced colleagues or manually sifting through digitized archives. This process could take anywhere from several hours to half a day, delaying repairs and increasing operational costs.

AI-Powered Search System Development

In collaboration with Recursive, a Japanese AI startup and official OpenAI partner, Kuken developed a search system integrated with a field app. The system uses AI models to extract and structure information from the company's extensive archive of drawings. Engineers can now search using part numbers, model names, or even screenshots, enabling quicker access to relevant documents. OpenAI's models were selected for their accuracy and cost-effectiveness, and the API's flexibility allowed seamless integration into the app.

Operational and Strategic Impact

The new system has transformed workflows at Kuken Industries. Engineers, regardless of experience level, can independently locate the necessary drawings, reducing reliance on senior colleagues. This not only accelerates repair processes but also allows newer engineers to focus on learning and problem-solving. Additionally, the system's ability to match replacement parts to original installations minimizes the need for custom manufacturing, further saving time and resources.

Future Plans and Implementation

The AI-powered system is currently being trialed at Kuken's Fukuoka factory, where it has already demonstrated significant time savings. Based on 2025 operational data, the company estimates a 90% reduction in annual man-hours dedicated to drawing searches. Kuken plans to refine the system further by improving its ability to interpret older, handwritten annotations and ensuring robust data security. Full-scale deployment across its Fukuoka and Chiba factories is targeted for late 2026.