Skip to content
Back to Blog
Smart Factory
2026-09-296 min read0

Turning Master Craftsmen's Know-How into Data — Korea's Manufacturing Tacit Knowledge AI Program Expands to KRW 290 Billion in 2027: A Preparation Roadmap for SMEs Facing Skilled Worker Retirement

Korea's program to turn skilled workers' judgment into AI-ready data is set to expand to KRW 290 billion in the 2027 budget proposal. This guide covers four steps SMEs can take before the call opens and the mistakes to avoid.

KITIM Consulting Team

What Tacit Knowledge Is and Why It Became a Policy Priority

Tacit knowledge is the know-how skilled workers build up over years on the job: techniques, judgment calls, and a feel for the work. Examples include judging a weld by the sound of the bead, reading temperature from the color of a casting, and sensing a bad fit through the resistance in their fingertips. This knowledge is hard to put into documents, so it has mostly been handed down through apprenticeship.

That model is breaking down. As the workforce ages and veteran workers retire, their know-how is at risk of disappearing. Meanwhile, reported figures put manufacturing AI adoption among Korean SMEs in the 0.1% range.

Earlier smart factory and AI programs focused mainly on sensor and equipment data. The tacit knowledge program is different because it aims to turn human judgment into data.

From the 2026 Pilot to the 2027 Full Program

  • 2026 pilot: Korea's trade and industry ministry allocated KRW 48 billion in supplementary budget to build tacit knowledge datasets and develop AI models across 30 selected projects (KRW 1.6 billion each), each led by a manufacturer–AI company consortium. An expert conference linked to the M.AX Alliance was held on June 12.
  • 2027 budget proposal: The proposal includes KRW 290 billion for the tacit knowledge program and KRW 115.8 billion for a National Manufacturing Data Library. Starting in 2027, the library is meant to build, store, and make manufacturing data available for use.
  • Caveat: These figures are still proposals that the National Assembly has not yet reviewed. The final amounts and support terms will depend on the budget passed in December and the calls for applications expected in early 2027.
  • Which Companies and Processes Are Best Positioned

  • Processes that produce unstructured data: welding, grinding and polishing, assembly, painting, heat-treatment judgment, and visual inspection, where video, audio, force/torque, and operator gaze data can be captured
  • Companies with hard-to-replace experts: root industries (casting, forging, welding and similar foundational processes), shipbuilding suppliers, and auto parts suppliers whose key experts are close to retirement
  • Companies with data infrastructure: firms that can link new data to existing MES or equipment data are likely to be rated more favorably during model development and validation.
  • Four Steps You Can Take Before the Call Opens

    Step 1: Map Your Tacit Knowledge

    For each process, list the decisions that only a particular person can make. Compare that list with your defect and rework history. This shows which judgments are tied directly to losses and tells you where to start.

    Step 2: Run a Data Collection Pilot

    Record work video synchronized with sensor data, and interview your experts so they explain their reasoning out loud. Before any filming, put consent procedures for personal data and image rights in place.

    Step 3: Settle Data Rights

    Agree in writing on data ownership, scope of use, and compensation among the individual workers, your company, and the AI developer. If you plan to contribute data to the national library, decide in advance how much you can share without exposing confidential know-how.

    Step 4: Explore Consortium Partners

    Work out how roles will be split among AI vendors, research institutes, and large buyer companies.

    Common Mistakes and Risks

  • Missing ground-truth labels: If you collect large volumes of video but never record what the expert actually decided (pass/fail, adjustment values, and so on), you cannot train a model on it.
  • Resistance from experts: Workers may fear that the project will replace them. Build in real incentives, such as compensation, formal master craftsman recognition, and roles as trainers.
  • Leakage of core know-how: Set up technology protection measures and agree on sharing limits before the project starts.
  • How KITIM Can Help

    KITIM assesses tacit knowledge process by process to set your priorities, then plans your application around them. We also map overlaps and links with AI-focused smart factory programs and M.AX-linked initiatives, and we manage your preparation timeline for the 2027 calls.

    If you are concerned about losing know-how as experienced workers retire, now is the time to prepare, before the call is announced. Request a consultation with KITIM to build a tacit knowledge AI strategy that fits your processes.

    Manufacturing Tacit KnowledgeSkilled Worker Know-HowManufacturing AIManufacturing Data LibraryM.AXSmart Factory2027 Budget
    매일 자동 업데이트

    이 분야 정부지원사업, AI가 찾아드립니다

    3분 기업진단만 완료하면 귀사에 맞는 공고를 적합도 점수와 함께 추천합니다. 무료입니다.

    AI 맞춤 공고 무료로 받기

    Need Consulting?

    Our technology innovation consultants will propose the optimal solution for your company.