[{"data":1,"prerenderedAt":572},["ShallowReactive",2],{"site-schema":3,"article-zh-tw-forward-deployed-engineer-fde-fdt":93,"article-related-zh-tw-forward-deployed-engineer-fde-fdt":556},{"@context":4,"@graph":5},"https:\u002F\u002Fschema.org",[6,72,84],{"@type":7,"@id":8,"name":9,"alternateName":10,"legalName":15,"foundingDate":16,"numberOfEmployees":17,"taxID":18,"url":19,"logo":20,"founder":21,"contactPoint":23,"address":37,"location":43,"sameAs":63},"Organization","https:\u002F\u002Fjoinx.co\u002F#organization","JoinX哲煜科技",[11,12,13,14],"JoinX","TWJOIN","哲煜科技","哲煜科技股份有限公司","JoinX Co., Ltd.","2016-04-28",80,"43774383","https:\u002F\u002Fjoinx.co","https:\u002F\u002Fjoinx.co\u002Fimages\u002Flogo-with-name.png",{"@id":22},"https:\u002F\u002Fjoinx.co\u002F#person-jack-lee",[24],{"@type":25,"contactType":26,"url":27,"telephone":28,"email":29,"areaServed":30,"availableLanguage":33},"ContactPoint","customer service","https:\u002F\u002Fjoinx.co\u002Fcontact-us","+886-2-8771-9095","service@joinx.co",[31,32],"TW","JP",[34,35,36],"zh-Hant","en","ja",{"@type":38,"streetAddress":39,"postalCode":40,"addressLocality":41,"addressRegion":42,"addressCountry":31},"PostalAddress","民生東路二段170號8樓","104","台北市中山區","台灣",[44,51,57],{"@type":45,"name":46,"address":47},"Place","JoinX 台中辦公室",{"@type":38,"streetAddress":48,"postalCode":49,"addressLocality":50,"addressRegion":42,"addressCountry":31},"台灣大道二段360號21樓C室","40453","台中市北區",{"@type":45,"name":52,"address":53},"JoinX 高雄辦公室",{"@type":38,"streetAddress":54,"postalCode":55,"addressLocality":56,"addressRegion":42,"addressCountry":31},"民族一路80號2樓之一(B4)","807","高雄市三民區",{"@type":45,"name":58,"address":59},"JoinX 東京辦公室",{"@type":38,"streetAddress":60,"postalCode":61,"addressLocality":62,"addressCountry":32},"東五反田5-22-37 Office Circle N 五反田 9樓","141-0022","東京都品川區",[64,65,66,67,68,69,70,71],"https:\u002F\u002Fwww.crunchbase.com\u002Forganization\u002Ftwjoin","https:\u002F\u002Fwww.facebook.com\u002FJoinX.TW","https:\u002F\u002Ftw.linkedin.com\u002Fcompany\u002F%E5%93%B2%E7%85%9C%E7%A7%91%E6%8A%80","https:\u002F\u002Fwww.youtube.com\u002Fc\u002Fjoinstuido%E5%93%B2%E7%85%9C%E7%A7%91%E6%8A%80","https:\u002F\u002Fmedium.com\u002Ftwjoin","https:\u002F\u002Fwww.104.com.tw\u002Fcompany\u002F1a2x6bjomb","https:\u002F\u002Fclutch.co\u002Fprofile\u002Ftwjoin","https:\u002F\u002Fwww.goodfirms.co\u002Fcompany\u002Ftwjoin",{"@type":73,"@id":22,"name":74,"alternateName":75,"jobTitle":76,"worksFor":77,"url":78,"sameAs":79},"Person","李秉哲","Jack Lee","創辦人暨執行長",{"@id":8},"https:\u002F\u002Fjack.joinx.co",[80,81,82,83],"https:\u002F\u002Fwww.instagram.com\u002Fjack88.podcast\u002F","https:\u002F\u002Fwww.threads.net\u002F@jack88.podcast","https:\u002F\u002Fjoinx.co\u002Fjack88\u002F","https:\u002F\u002Fwww.linkedin.com\u002Fin\u002Fjacklee-taiwan\u002F",{"@type":85,"@id":86,"url":19,"name":9,"alternateName":87,"publisher":88,"inLanguage":89},"WebSite","https:\u002F\u002Fjoinx.co\u002F#website",[11,13,9],{"@id":8},[90,91,92],"zh-Hant-TW","en-US","ja-JP",{"id":94,"title":95,"body":96,"category":524,"cover":525,"dateModified":526,"description":527,"extension":528,"faq":526,"hasCoverTitle":529,"locale":530,"meta":531,"navigation":532,"need":533,"path":534,"seo":535,"seoDescription":526,"seoTitle":526,"sources":536,"stem":546,"takeaways":547,"time":553,"type":554,"__hash__":555},"content\u002Fzh-tw\u002Farticle\u002Fforward-deployed-engineer-fde-fdt.md","FDE 是什麼？Forward Deployed Engineer 和 FDT、AI 顧問、外包怎麼選（2026）",{"type":97,"value":98,"toc":506},"minimark",[99,107,112,115,118,122,125,128,132,135,232,241,245,248,251,285,289,292,295,372,375,378,407,419,422,426,429,467,470,475,478,482,485,489,492,496,499,503],[100,101,102,106],"p",{},[103,104,105],"strong",{},"FDE（Forward Deployed Engineer）是派工程師進到客戶的流程裡，邊了解業務邊開發的做法；FDT（Forward Deployed Team）則是派一整個團隊。"," 兩者交付的都是能用的功能，而不是建議書。企業 AI 導入卡住的地方，常常是題目找不準、做出來沒人用，這正是「進場」做法要解決的問題。",[108,109,111],"h2",{"id":110},"fdeforward-deployed-engineer是什麼","FDE（Forward Deployed Engineer）是什麼？",[100,113,114],{},"FDE 是派到客戶端的工程師，直接看客戶的資料與流程，在現場把軟體調整成客戶能用的樣子。這個職稱最早由美國資料分析公司 Palantir 帶起，近年 OpenAI 等 AI 公司也設立 FDE，協助企業把模型接進實際工作。",[100,116,117],{},"和一般工程師不同，FDE 的工作從理解業務開始：先弄清楚第一線人員怎麼做事、資料在哪裡，再決定要開發什麼。FDE 通常同時負責和客戶溝通、寫程式與部署，所以很重視工程師對業務的判斷力。",[108,119,121],{"id":120},"為什麼企業-ai-導入需要進場的做法","為什麼企業 AI 導入需要「進場」的做法？",[100,123,124],{},"因為 AI 題目通常要用真實資料試過，才知道做不做得到、值不值得做。只靠會議與文件定規格，常常做完才發現資料拿不到、準確率不夠，或第一線人員根本不會用。",[100,126,127],{},"一般軟體可以先把規格寫清楚再開發，AI 專案比較難。模型的表現取決於資料品質，流程裡哪一步交給 AI、哪一步要由人確認，也要實際試過才知道。進場的做法讓開發的人直接看到資料與工作現場，能更早發現問題、調整方向。NIST 的 AI 風險管理框架也把「了解使用情境」與「人為監督」列為治理重點，這兩件事都要靠貼近現場才做得好。",[108,129,131],{"id":130},"fdefdtai-顧問一般外包差在哪","FDE、FDT、AI 顧問、一般外包差在哪？",[100,133,134],{},"差別在誰進場、交付什麼，以及做完之後誰負責。AI 顧問交付建議與規劃，FDE 交付功能，FDT 交付從原型到上線的整套系統，一般外包則依照規格開發。",[136,137,138,160],"table",{},[139,140,141],"thead",{},[142,143,144,148,151,154,157],"tr",{},[145,146,147],"th",{},"做法",[145,149,150],{},"誰進場",[145,152,153],{},"交付什麼",[145,155,156],{},"適合",[145,158,159],{},"要注意",[161,162,163,181,198,215],"tbody",{},[142,164,165,169,172,175,178],{},[166,167,168],"td",{},"AI 顧問",[166,170,171],{},"顧問",[166,173,174],{},"評估報告、導入規劃、培訓課程",[166,176,177],{},"還在評估方向、要先有整體規劃",[166,179,180],{},"實作要另外找團隊，建議和實作容易斷在中間",[142,182,183,186,189,192,195],{},[166,184,185],{},"FDE",[166,187,188],{},"一位或少數幾位工程師",[166,190,191],{},"在現場邊了解邊開發的功能",[166,193,194],{},"已有產品或平台，要客製接進客戶流程",[166,196,197],{},"設計、專案管理與上線後維運多半要企業自己補",[142,199,200,203,206,209,212],{},[166,201,202],{},"FDT",[166,204,205],{},"PM、設計、前後端與 AI 工程師",[166,207,208],{},"PoC 或 MVP，接著正式上線與維運",[166,210,211],{},"題目還要一起找，做出來還要長期用",[166,213,214],{},"要有內部窗口一起參與，題目才找得準",[142,216,217,220,223,226,229],{},[166,218,219],{},"一般外包",[166,221,222],{},"依規格開發的團隊",[166,224,225],{},"照規格做完的系統",[166,227,228],{},"規格已經清楚、需求不太會變",[166,230,231],{},"AI 題目常常做了才知道要改，規格很難一次寫對",[100,233,234,235,240],{},"這四種做法沒有絕對的好壞。如果企業內部已經有 PM 與設計，只缺能寫 AI 的工程師，FDE 就夠了；如果連題目都還要一起找，做出來也要有人長期維護，FDT 會比較完整。AI 顧問的角色可以參考",[236,237,239],"a",{"href":238},"\u002Farticle\u002F2026-ai-enterprise-poc-ai-guide\u002F","AI 顧問是什麼","。",[108,242,244],{"id":243},"什麼情況適合找-fdt","什麼情況適合找 FDT？",[100,246,247],{},"當企業還沒想清楚 AI 要用在哪裡、內部也沒有完整的開發團隊，而做出來的系統要接進既有流程並長期使用時，就適合 FDT。只要其中一項不成立，其他做法可能更省。",[100,249,250],{},"以下幾個訊號，代表 FDT 比較適合：",[252,253,254,261,267,273,279],"ol",{},[255,256,257,260],"li",{},[103,258,259],{},"題目還沒定。"," 知道某個流程很花人力，但不確定 AI 能幫上哪一段。",[255,262,263,266],{},[103,264,265],{},"內部沒有開發團隊。"," 沒有 PM、設計或工程師可以接手顧問的建議。",[255,268,269,272],{},[103,270,271],{},"要接既有系統。"," AI 要讀 ERP、CRM、訂單或文件資料，不是獨立的工具。",[255,274,275,278],{},[103,276,277],{},"上線後要長期用。"," 系統會隨業務調整，需要有人持續維運。",[255,280,281,284],{},[103,282,283],{},"希望先看到成果再投入。"," 想先用小範圍驗證，確定有用再擴大預算。",[108,286,288],{"id":287},"fdt-怎麼分階段合作與驗收","FDT 怎麼分階段合作與驗收？",[100,290,291],{},"建議分成陪跑與執行兩個階段收費。陪跑階段做出 PoC 或 MVP，依事先定好的門檻驗收；通過了，再由同一組人進入執行階段，把原型做成正式上線的系統。",[100,293,294],{},"依 JoinX 整理的台灣客製專案行情，各階段的費用與時程如下：",[136,296,297,314],{},[139,298,299],{},[142,300,301,304,308,311],{},[145,302,303],{},"階段",[145,305,307],{"align":306},"right","預估費用",[145,309,310],{},"常見內容",[145,312,313],{},"參考時程",[161,315,316,330,344,358],{},[142,317,318,321,324,327],{},[166,319,320],{},"陪跑階段：PoC",[166,322,323],{"align":306},"NT$30 萬～80 萬",[166,325,326],{},"一個場景、有限資料、人工監督下測試",[166,328,329],{},"4～8 週",[142,331,332,335,338,341],{},[166,333,334],{},"陪跑階段：MVP",[166,336,337],{"align":306},"NT$50 萬～120 萬",[166,339,340],{},"做出可以給真實使用者操作的最小版本",[166,342,343],{},"6～12 週",[142,345,346,349,352,355],{},[166,347,348],{},"執行階段：正式上線",[166,350,351],{"align":306},"NT$80 萬～250 萬",[166,353,354],{},"權限、系統串接、監控、測試與部署",[166,356,357],{},"2～4 個月",[142,359,360,363,366,369],{},[166,361,362],{},"跨部門、多系統",[166,364,365],{"align":306},"NT$250 萬以上",[166,367,368],{},"多個場景或 AI Agent、完整治理",[166,370,371],{},"4 個月以上",[100,373,374],{},"AI 模型 API 與雲端費用依用量另計，上線後的維運另訂合約。內部培訓建議包含在各階段的服務裡，讓同仁在陪跑階段就開始使用，而不是上線後才另外上課。",[100,376,377],{},"以 JoinX 做過的專案為例：",[379,380,381,387,397],"ul",{},[255,382,383,386],{},[103,384,385],{},"壽險公司的 AI 教練系統 Demo："," 約三週做出可以實際操作的展示版本，讓內部用成果評估要不要往下做。這就是陪跑階段的典型做法。",[255,388,389,392,393,240],{},[103,390,391],{},"聯經數位的有聲書平台："," 開發前先定好驗收門檻，角色辨識成功率 90%、文本拆解成功率 75% 以上，再把整條製作流程自動化，平台後來獲得 iF 設計獎。完整做法見",[236,394,396],{"href":395},"\u002Fcase-studies\u002Faicast\u002F","聯經數位案例",[255,398,399,402,403,240],{},[103,400,401],{},"國家原子能科技研究院的渦電流訊號判讀："," 先單頻道、後多頻道分階段開發，AI 辨識與自動化控制拆成兩套，從 2022 年合作到 2025 年。完整做法見",[236,404,406],{"href":405},"\u002Fcase-studies\u002Fnari\u002F","國家原子能科技研究院案例",[100,408,409,410,414,415,240],{},"更多案例與費用區間，可以參考",[236,411,413],{"href":412},"\u002Farticle\u002Ftaiwan-enterprise-ai-adoption-cases-2026\u002F","企業 AI 導入案例","；JoinX 的分階段合作方式，見",[236,416,418],{"href":417},"\u002Fai-implementation-team\u002F","AI 導入陪跑團隊",[420,421],"article-cta",{},[108,423,425],{"id":424},"選-fde-或-fdt-夥伴時要問哪些問題","選 FDE 或 FDT 夥伴時，要問哪些問題？",[100,427,428],{},"要問清楚進場的是誰、怎麼驗收，以及做完之後誰負責。這三件事決定建議和實作會不會斷在中間，也決定系統上線後有沒有人能接著改。",[252,430,431,437,443,449,455,461],{},[255,432,433,436],{},[103,434,435],{},"進場的是哪些角色？"," 只有工程師，還是包含 PM、設計與 AI 工程師？",[255,438,439,442],{},[103,440,441],{},"陪跑階段怎麼驗收？"," 門檻是準確率、處理時間，還是使用者能完成某個任務？達不到門檻時怎麼處理？",[255,444,445,448],{},[103,446,447],{},"做過哪些已經上線的 AI 專案？"," 只有 Demo，還是有正式在用、有人維運的系統？",[255,450,451,454],{},[103,452,453],{},"培訓怎麼安排？"," 是含在服務裡，還是另外開課收費？",[255,456,457,460],{},[103,458,459],{},"上線後誰維運？"," 同一個團隊，還是要另外找人接手？別家做的系統能不能接？",[255,462,463,466],{},[103,464,465],{},"資料怎麼保護？"," 使用的 AI 服務會不會拿資料去訓練？機敏資料怎麼分級？",[108,468,469],{"id":469},"常見問題",[471,472,474],"h3",{"id":473},"fde-和一般的駐點工程師有什麼不同","FDE 和一般的駐點工程師有什麼不同？",[100,476,477],{},"駐點工程師多半依照客戶的指示做事，FDE 則要自己判斷該開發什麼。FDE 會先了解業務與資料，再決定功能與做法，所以很重視對業務的理解，不只是寫程式的能力。",[471,479,481],{"id":480},"中小企業也適合找-fde-或-fdt-嗎","中小企業也適合找 FDE 或 FDT 嗎？",[100,483,484],{},"適合，但建議從一個場景開始。先挑一個資料拿得到、結果能由人複核的流程做 PoC，驗收通過再擴大，預算比較好控制。",[471,486,488],{"id":487},"fdt-和-ai-顧問可以一起用嗎","FDT 和 AI 顧問可以一起用嗎？",[100,490,491],{},"可以，但常常不需要。FDT 的陪跑階段本身就包含找題目與評估；如果企業已經請顧問做完規劃，FDT 可以直接依規劃做 PoC，省下重複評估的時間。",[471,493,495],{"id":494},"陪跑階段做出來的-poc可以直接上線嗎","陪跑階段做出來的 PoC，可以直接上線嗎？",[100,497,498],{},"通常不行。PoC 是在有限資料與人工監督下驗證可行性，正式上線還要補齊權限、系統串接、監控與測試。由同一組人接著做，可以沿用 PoC 的成果，不用重新交接。",[471,500,502],{"id":501},"ai-導入一定要派人到公司現場嗎","AI 導入一定要派人到公司現場嗎？",[100,504,505],{},"不一定。重點是開發的人要直接接觸資料與第一線人員，訪談與測試可以到現場，日常開發也可以遠端進行。關鍵在於能不能快速拿到回饋，而不是人一直待在現場。",{"title":507,"searchDepth":508,"depth":508,"links":509},"",2,[510,511,512,513,514,515,516],{"id":110,"depth":508,"text":111},{"id":120,"depth":508,"text":121},{"id":130,"depth":508,"text":131},{"id":243,"depth":508,"text":244},{"id":287,"depth":508,"text":288},{"id":424,"depth":508,"text":425},{"id":469,"depth":508,"text":469,"children":517},[518,520,521,522,523],{"id":473,"depth":519,"text":474},3,{"id":480,"depth":519,"text":481},{"id":487,"depth":519,"text":488},{"id":494,"depth":519,"text":495},{"id":501,"depth":519,"text":502},"技術分享","\u002Fimages\u002Fblog\u002Fforward-deployed-engineer-fde-fdt.webp",null,"FDE（Forward Deployed Engineer）是派工程師進到客戶的流程裡，邊了解業務邊開發的做法，近年被 AI 公司用來協助企業把模型接進實際工作。FDT（Forward Deployed Team）則是把這個做法擴大成一整個團隊，由 PM、設計、前後端與 AI 工程師從找題目做到上線。本文比較 FDE、FDT、AI 顧問與一般外包的差異，說明什麼情況適合哪一種，以及分階段合作與驗收的方式。","md",false,"zh-tw",{},true,"AI 導入","\u002Fzh-tw\u002Farticle\u002Fforward-deployed-engineer-fde-fdt",{"title":95,"description":527},[537,540,543],{"name":538,"issuer":539},"Forward Deployed Software Engineer 職務說明","Palantir Technologies",{"name":541,"issuer":542},"Forward Deployed Engineer 職務說明","OpenAI",{"name":544,"issuer":545},"AI Risk Management Framework（AI RMF 1.0）","美國國家標準與技術研究院（NIST）","zh-tw\u002Farticle\u002Fforward-deployed-engineer-fde-fdt",[548,549,550,551,552],"FDE 是派工程師進到客戶的流程裡邊了解邊開發，交付的是能用的功能，不是建議書。","FDT 把 FDE 擴大成一整個團隊，PM、設計、前後端與 AI 工程師都在裡面，從找題目做到上線與維運。","AI 顧問適合還在評估方向的階段，一般外包適合規格已經清楚的系統，AI 題目常常要做了才知道怎麼改。","分階段合作比較穩：先用 PoC 或 MVP 依事先定好的門檻驗收，通過了再投入正式上線的預算。","依 JoinX 的台灣客製專案行情，PoC 約 NT$30 萬至 80 萬，MVP 約 NT$50 萬至 120 萬，正式上線約 NT$80 萬至 250 萬。","2026\u002F11\u002F04","blog","sJdWAFQsGBbwnWfQCvFGlcsFdKToindcywL434ZDFAA",[557,562,567],{"slug":558,"title":559,"time":560,"cover":561,"need":533},"ai-adoption-effectiveness-exception-handling","AI 導入後如何判斷成效？從流程改善到例外處理","2026\u002F07\u002F31","\u002Fimages\u002Fblog\u002Fai-adoption-effectiveness-exception-handling.webp",{"slug":563,"title":564,"time":565,"cover":566,"need":533},"ai-adoption-failure-patterns","AI 導入為什麼失敗？五種典型樣態與救援實務","2026\u002F07\u002F29","\u002Fimages\u002Fblog\u002Fai-adoption-failure-patterns.webp",{"slug":568,"title":569,"time":570,"cover":571,"need":533},"how-to-start-ai-adoption","AI 導入該怎麼開始？AI 顧問如何協助企業評估 AI 優先順序","2026\u002F07\u002F27","\u002Fimages\u002Fblog\u002Fhow-to-start-ai-adoption.webp",1793765008199]