KOREA VIDEO REVIEW · NATIONAL PLATFORM
A State-Built AI Manufacturing Platform: Korea's KAMP — Ideal, Reality and Upgrade
01What this video is about
This March 2026 video from KTV (Korea's policy broadcaster) is formatted as a "policy fact check": the trigger was media reports that KAMP, the AI manufacturing platform the government funded with billions of won, had a low usage record; the Ministry of SMEs and Startups (중소벤처기업부) used the video to respond — it will upgrade KAMP into the "제조AI 24 (Manufacturing AI 24)" platform, improve the collection, sharing and linking system for SMEs' manufacturing data, and study offering incentives for data sharing[1].
The government's own channel addressing the "low usage" criticism head-on is honesty worth commending — and it makes this video excellent material for observing the national-platform model.
02What KAMP is: the national team's manufacturing-AI resource pack
KAMP went live in 2020, with a design logic that centrally supplies the three things small and medium factories lack for doing AI[2]:
- Data: public manufacturing-AI datasets (machining, inspection, equipment data), giving factories without their own data a starting point;
- Compute and tools: a cloud development environment, sparing them from building their own infrastructure;
- Experts: matching with AI service providers and consultants, filling SMEs' talent gap.
Academically this prescribes exactly for manufacturing ML's three big bottlenecks (data, compute, cross-domain talent)[3] — the design is right. It is also the archetype of Korea's "국가가 깔아주는 (the state paves the road)" model: the same problem as Germany's Mittelstand-Digital guidance network (see the Germany installment) solved differently — Germany gives guidance, Korea gives a platform.
03The usage debate: the blind spots of the platform model
So why get questioned on usage? The officials' own list of countermeasures in the video is the flip side of the answer: to "improve data collection, sharing and linking" and to "offer sharing incentives"[1] — in other words, these are exactly what got stuck:
- The factory side has no data to upload: however good the platform, a factory whose drawings are still on paper and whose machines are not networked doesn't even have the ticket to participate — the first layer of CPS layered theory (digitization) isn't passed, so the upper-layer services can't be used[4].
- Insufficient sharing incentives: manufacturing data involves process secrets, so SMEs are naturally wary of "handing over the data" — this is not a technical problem but a matter of trust and incentive design.
- Scarce problem-definition ability: the platform waits for you to bring a problem, but most factories cannot articulate "what I want AI to do for me" — precisely the thing, discussed in the Germany installment, that guidance is harder to supply than a platform.
To be fair: these blind spots are not unique to KAMP; they are shared challenges of every "build the platform first and wait for people to come" model. Korea's upgrade direction (data system plus incentives) is aimed accurately enough.
04The conclusion for Taiwan shops: don't wait for the platform, prepare your data first
Taiwan also has policy resources related to smart machinery and AI. KAMP's experience gives Taiwan shops three pragmatic conclusions:
- Use policy resources, but bring your own ticket: whether it's a subsidy, a platform or institutional guidance, the precondition for catching the resource is that your data is already digitized and your problem already defined — two things no platform can do for you.
- Grow data out of daily workflow: the most sustainable digitization is not project-style "uploading data" but making data a by-product of daily work — a preparation chain where AI reads 2D drawings, models, generates code, simulates and verifies leaves behind structured drawing, operation and verification data for every job (see the complete guide to CNC auto-programming) — and that is your data footing for plugging into any platform or subsidy program in the future.
- Address confidentiality concerns head-on: KAMP got stuck on trust in data sharing, and Taiwan shops should likewise ask about data terms when choosing tools — BestAI CAM's commercial terms are explicit: confidential drawings are not retained and not used for model training (for the security framework, see the on-premise and security checklist).
Between the ideal and the reality of a national platform lies each factory's own degree of digitization — that is the lesson most worth taking away from this KAMP video.
05FAQ
What is KAMP, and how effective has it been?
An AI manufacturing platform set up by Korea's Ministry of SMEs and Startups in 2020: it provides manufacturing datasets, a cloud development environment and expert matching to help small and medium factories without an AI team develop manufacturing AI. In 2026 media questioned its low usage record, and officials responded that it will be upgraded to the "Manufacturing AI 24" platform with an improved data collection and sharing system — the policy design is recognized, but the landing is stuck on how digitized the factory side's data is and on incentives to share.
Why would a government platform suffer low usage?
Three structural blind spots: most factories have not digitized their own data (they don't even have the ticket to participate), manufacturing data involves process secrets so willingness to share is low, and factories lack the ability to "define a pain point as an AI problem." These are shared challenges of every "build the platform first and wait for people to come" model, not unique to Korea.
How can a Taiwan shop make good use of similar policy resources?
Bring your own ticket first: digitize your drawings and processes (accumulate them as by-products of daily workflow, such as the structured job data left behind by AI reading drawings and generating code), and define pain points as measurable goals. With data and a problem definition in hand, subsidy programs and institutional guidance can actually catch you; also watch the data terms — ask clearly about retention and training-use policies for confidential drawings.
Get notified when new articles and video guides go live — no inbox flooding, one-click unsubscribe. (newsletter in Chinese)
DATA FIRST, PLATFORM LATER
Don't wait for the platform to pave the road — turn your drawings into digital assets first
AI reads drawings, models, generates code and verifies, and job data accumulates automatically; confidential drawings are not retained and not used for training — data sovereignty stays in your hands.
Contact an onboarding advisor Training courses📋 Download the free "CNC Outsourcing Checklist" (printable) (中文)
06References
- KTV 국민방송 (YouTube). 중소벤처기업부 "제조데이터 수집·공유·연계 체계 개선해 AI 제조플랫폼 활성화 노력" (published 2026-03-06). youtube.com/watch?v=A5A66O0E-E4
- 중소벤처기업부 (YouTube). KAMP 플랫폼에서 무엇을 얻을 수 있는 지 알아본다! [마이스터백2]. youtube.com/watch?v=0PSwcvRj_iU
- Wuest, T., Weimer, D., Irgens, C., & Thoben, K.-D. (2016). Machine learning in manufacturing: advantages, challenges, and applications. Production & Manufacturing Research, 4(1), 23–45.
- Lee, J., Bagheri, B., & Kao, H.-A. (2015). A Cyber-Physical Systems architecture for Industry 4.0-based manufacturing systems. Manufacturing Letters, 3, 18–23.