In many Japanese factories, some of the most valuable information has never been properly written down.

An experienced technician may recognise a defective component from an unusual vibration. A quality manager may know which small design change could create problems months later. Another employee may remember that a similar failure occurred ten years ago—but only because they personally investigated it.

Japan calls this kind of experience 暗黙知 (anmokuchi), or tacit knowledge: information acquired through practice that is difficult to express in a manual.

Tokyo startup JINGS is developing an AI system intended to capture and organise that knowledge before it disappears.

JINGS has opened an alpha test of Lean AI and is inviting manufacturing companies to participate.

According to the company, the system can analyse factory videos, technical documents and interviews with experienced employees. It then helps organise the reasoning behind operational decisions: what the worker noticed, why it mattered and which action followed.

The objective is not simply to create another internal chatbot.

Consider a veteran quality inspector who rejects a component because its surface “does not look right.” The final decision may be recorded, but the visual clues, comparisons and accumulated experience behind it often remain inside that employee’s head.

Lean AI is intended to help extract those criteria and transform them into knowledge that can support training, process improvement and standardisation.

Japanese manufacturers have accumulated decades of reports, inspection records, spreadsheets and technical documents. However, finding useful information inside them can be surprisingly difficult.

Files may use different formats. The same defect may have several names. Reports can be distributed across departments, factories and individual computers. Important lessons from previous failures—often called 過去トラ (kako-tora), or past trouble cases—may never reach the engineers designing the next product.

JINGS already develops AI tools for quality assurance, production engineering and industrial design. Its systems can search previous failure reports, identify related documents and propose issues that engineers may need to examine during a design or process review.

The company says its AI is designed to assist human judgement, not replace it. That distinction matters in manufacturing, where a confident but incorrect AI answer could result in defective products, expensive recalls or safety problems.

Employees must therefore be able to inspect the evidence behind a suggestion and make the final decision themselves.

JINGS was established in 2024 and is led by CEO Haruka Mikami. The company has been recognised as a startup originating from the University of Tokyo’s Matsuo Laboratory ecosystem.

Its disclosed work includes projects involving automotive suppliers and manufacturers such as Glory, Sysmex and Astemo. JINGS says it can begin with information companies already possess, including Excel files, handwritten inspection records, design documents and failure reports.

That could make the technology more accessible than projects requiring factories to install an entirely new network of sensors or replace their existing systems.

However, the alpha test will need to demonstrate that Lean AI can understand knowledge that is incomplete, inconsistent and highly specific to an individual factory.

Recording an interview is relatively easy. Converting an expert’s explanations into reliable criteria that another employee can apply is much harder.

JINGS is currently looking for manufacturing companies willing to test the platform. It has not publicly specified a deadline, participation price, technical requirements or the number of companies it will accept.

The most important results will not be the amount of information processed. The real questions are whether Lean AI can reduce the time needed to investigate problems, improve employee training and prevent knowledge from disappearing when experienced workers leave.

Japan does not lack industrial information. Much of it is simply trapped in old documents, isolated departments and the memories of people approaching retirement.

If JINGS can make that knowledge reusable without removing human verification, Lean AI could become more than another generative-AI product. It could help preserve part of the operational memory on which Japanese manufacturing was built.

SAKIME
East Asian startups, technology and international investment opportunities.

Web: sakime.jp
Author — Adrià Mas Rodríguez

This article is provided for informational purposes and does not constitute an investment recommendation or solicitation.