AI·머신러닝 · LLM기타 산업 (public sector construction supervision)
서울시설공단이 26년간 쌓인 공사감독 경험 1만1,835건을 AI 업무지원 플랫폼으로 만들었다.
출처 [1] 인공지능신문 · 2026년 10월 23일 · “서울시설공단, 1만1천여 건 공사감독 경험 AI로 구현…건설현장 안전관리 지원” ↗
도입 단계: 시범(PoC) — A named organisation is trialling it on real data, bounded scope or time.
요약
서울시설공단이 26년간 쌓인 공사감독 경험 1만1,835건을 AI 업무지원 플랫폼으로 만들었다. 단계별 절차와 점검 항목을 안내하고 서류 누락·오류를 AI가 미리 걸러, 신입 감독자와 소규모 시공사가 같은 기준으로 일하게 하는 것이 목적이다.
Kenny의 관점
선사에도 26년치 노하우를 머릿속에만 가진 사람들이 있다. 이 사례의 가치는 AI가 아니라 1만 건 넘는 감독 기록을 기준으로 정리해 둔 데 있다. 해운에서 같은 일을 하려면 먼저 사고·클레임 기록을 이렇게 정리해야 한다. 실제 현장 사용 수치가 나오지 않으면 이것은 아직 기록 정리 사업이다.
AI 초안 · Kenny 검토 승인
분석 (영문)
Stage: pilot — a named organisation has built and is putting the tool to work, but the article gives no usage scale or results, so production cannot be claimed. Pattern: capturing a retiring expert base into an assistant that standardises judgment — the "preserve expert judgment" half of thesis 1.
Inference: the shape — a long internal case record turned into stage-gated checklists plus document pre-checks — maps onto port and shipping operations documentation (B/L checks, terminal procedures) where knowledge also sits with a shrinking group of senior staff. The article says nothing on how wrong guidance is caught or who is accountable when the AI misses a document error.
원문에서 확인한 사실 (영문)
- F1Seoul Facilities Corporation turned 11,835 construction-supervision cases accumulated over 26 years into an AI-based work-support platform, which it developed in-house. [1]
- F2When construction information is entered, the platform guides stage-by-stage tasks and checkpoints from pre-construction through construction to completion, and AI reviews prepared administrative documents for missing items or errors; information is accessible on smartphones at the site. [1]
- F3The stated focus is standardising supervisory judgments that can vary with a supervisor's experience and skill, so that new supervisors and small contractors manage quality and safety to consistent criteria. [1]
- F4The corporation plans to share its platform-building know-how with local governments and contractors nationwide through customised guides, and to publish a "Go-Stop" casebook of good practice, recurring errors, findings and disputes. [1]
언급된 조직: Seoul Facilities Corporation
반론 — Codex 검증 메모
Codex가 각 문장을 원문과 대조하고 반대 증거를 찾은 기록입니다 (영문 원문). 판정: Codex 검증 · 부분 일치.
- 반대 증거The corporation's supervision-performance table reports 13,133 cumulative cases as of February 2026. It does not specify whether that total and the article's 11,835 platform cases have the same scope, so it is not a demonstrated correction. [2]
문장별 검증 메모 7건
- analysispartly-supported: The named organisation is putting the tool to work. — The article reports development and describes intended functions, but does not explicitly document a live trial or day-to-day use.
- analysisunsupported: The platform captures a retiring expert base. — The article describes accumulated expertise but says nothing about retirement.
- takeunsupported: 선사에도 26년치 노하우를 머릿속에만 가진 사람들이 있다. — The source does not establish this claim about shipping companies or people with 26 years of knowledge.
- takeunsupported: Every carrier has people holding 26 years of know-how in their heads. — The source does not establish this universal claim. The Korean Take makes the narrower claim that such people exist at shipping companies.
- takepartly-supported: 1만 건 넘는 감독 기록을 기준으로 정리해 둔 데 — The article says accumulated procedures and criteria were reflected in the platform; it does not say the more than 10,000 case records were first organised into criteria.
- takepartly-supported: The 11,000-plus supervision records were organised into criteria first. — The article does not describe that preparation step or its sequence.
- stagepilot → ?: The source names the developer and describes the platform, but gives no explicit trial on real data, bounded trial scope or period, or routine operational use. None of the taxonomy's stage rules is established clearly enough to assign a replacement.
출처
- [1] 인공지능신문 — “서울시설공단, 1만1천여 건 공사감독 경험 AI로 구현…건설현장 안전관리 지원” 원문
- [2] sisul.or.kr 반대 증거 (Codex)
이어 읽기
Stay in the loop.