把二十年的工地判斷、變成一座可傳承的知識庫
Turning 20 Years of Site Judgment Into a Transmissible Knowledge Base
🇹🇼 中文
一句話
裝修這行最值錢的東西、是老師傅腦袋裡那些「現場才學得到」的判斷——而它們大多沒被寫下來、人一退休就帶走了。我用 AI 當訪談者、把這些經驗一條一條挖出來、查證、結構化、變成一座留得住、傳得下去的知識庫。
問題
工地的知識、幾乎都是「默會知識」:磁磚怎麼驗、天花板為什麼會裂、矽利康哪裡不能用、這面牆能不能動……這些不在課本上、也 google 不到完整答案、因為會的人不寫、寫的人不會。
這帶來兩個問題:對外、屋主被資訊落差坑;對內、這身經驗無法被盤點、無法被傳承、也無法變成任何產品的底料。
怎麼用 AI 落地
我發明了一套「我帶判斷、AI 帶結構」的訪談法:
- AI 提大方向問題、逼我把腦袋裡那些「想當然耳」的東西講清楚;
- 我補現場細節與判斷、那個「外行以為 X、業內其實 Y」的落差、就是最值錢的礦;
- 凡涉及物性、法規、數字、一律查證——引原廠規格、國家標準(CNS)、法條、不靠記憶、不腦補;
- 結構化成知識庫、再讓它往下游長:變成對外的科普短影音、變成工具 app 的底層邏輯。
特別要守的紀律:不打臉同行、不把經驗值講成絕對、通俗講法不說它錯——業內人的可信度、來自分寸。
量化結果
- 一座七大類(磁磚 / 防水 / 板材 / 天花板 / 水電 / 矽利康 / 營造管理)的第一手工法與驗收標準知識庫、持續累積。
- 已餵出對外科普短影音(其中一支單平台觀看破 500)、也成為工具 app 的計算與驗收邏輯來源。
- 把原本「只在我腦袋、只在工地」的東西、變成可被搜尋、可被引用、可被傳承的資產。
這個案例證明什麼
這是我整個故事的核心命題的實證:我的二十年經驗是原料、AI 是讓它升級的工具。
AI 對領域專家最大的價值、不是取代你的判斷、是幫你把判斷萃取出來、結構化、再放大。把最不數位的工地經驗、變成數位資產——這件事、沒有那二十年的底、AI 自己做不到;沒有 AI、那二十年也只能爛在腦袋裡。兩個加起來、才是護城河。
🇬🇧 English
One-liner
The most valuable thing in renovation is the judgment a veteran carries in his head — the kind you can only learn on site. Most of it was never written down, and walks out the door when someone retires. I used AI as an interviewer to dig that experience out piece by piece, fact-check it, structure it, and turn it into a knowledge base that stays — and can be passed on.
The Problem
Site knowledge is almost all "tacit": how to inspect tiling, why a ceiling cracks, where silicone must not be used, whether a wall can be removed… None of it is in textbooks, and you can't fully google it — because those who know don't write, and those who write don't know.
That creates two problems: externally, homeowners get burned by the information gap; internally, this body of experience can't be inventoried, can't be passed on, and can't become the foundation of any product.
How I Landed AI on It
I built an interview method — I bring the judgment, AI brings the structure:
- AI asks the big-picture questions, forcing me to spell out the things I'd taken for granted;
- I add the on-site detail and judgment — and that gap ("outsiders assume X; insiders know it's actually Y") is the richest ore;
- Anything touching material properties, regulations, or numbers gets verified — citing manufacturer specs, national standards (CNS), and statutes; no relying on memory, no making things up;
- Structured into a knowledge base, then let to grow downstream: into public-education short videos, and into the underlying logic of tool apps.
A discipline I hold strictly: never embarrass peers, never present rules-of-thumb as absolutes, never call common usage "wrong" — an insider's credibility comes from knowing the measure of things.
Measurable Results
- A first-hand knowledge base across seven categories (tiling / waterproofing / boards / ceilings / electrical & plumbing / silicone / construction management), continuously growing.
- Already feeds public-education short videos (one passed 500 views on a single platform) and serves as the calculation and inspection logic behind tool apps.
- Turned what "lived only in my head, only on site" into an asset that is searchable, citable, and transmissible.
What This Case Proves
This is the proof of my whole story's core thesis: my twenty years of experience are the raw material; AI is the tool that upgrades it.
AI's greatest value to a domain expert isn't replacing your judgment — it's helping you extract, structure, and amplify it. Turning the least digital site experience into a digital asset: AI can't do it without those twenty years, and those twenty years would rot in someone's head without AI. Together, that's the moat.