{
  "source": "FDE模式行业观察与实践",
  "pagesTotal": 83,
  "coverage": "全部页面均提供主题和摘要索引；不发布逐字正文、原图或可还原报告版式的完整转录。",
  "pages": [
    {"page":1,"kind":"封面","section":"报告信息","topics":["报告标题","出品机构"],"summary":"报告封面，说明主题为 AI 落地实践中的 FDE 模式，由腾讯研究院出品。","sourceRef":"report:p1","fullTextPublished":false},
    {"page":2,"kind":"署名","section":"报告信息","topics":["顾问","作者","致谢"],"summary":"列出报告顾问、主笔和参与调研支持的产品、行业、云服务及生态团队。","sourceRef":"report:p2","fullTextPublished":false},
    {"page":3,"kind":"正文","section":"报告定位","topics":["研究对象","研究问题","表达边界"],"summary":"界定报告面向企业客户与生态伙伴，研究 FDE 的价值、机制、规模化条件和平台厂商实践边界。","sourceRef":"report:p3","fullTextPublished":false},
    {"page":4,"kind":"正文","section":"核心结论","topics":["组织采用","生产落地","前线学习","资产沉淀"],"summary":"归纳 AI 落地从能力展示转向组织使用，并强调生产交付与可复用知识沉淀必须同时发生。","sourceRef":"report:p4","fullTextPublished":false},
    {"page":5,"kind":"目录","section":"报告目录","topics":["章节结构","附件结构"],"summary":"列出十章正文、研究方法附录、国内外招聘数据附件和参考文献的整体结构。","sourceRef":"report:p5","fullTextPublished":false},
    {"page":6,"kind":"前言","section":"报告前言","topics":["AI 生产落地","角色误解","研究目的"],"summary":"说明企业关注点已从模型能力转向真实系统与业务结果，并澄清 FDE 不等于驻场或传统实施。","sourceRef":"report:p6","fullTextPublished":false},
    {"page":7,"kind":"章节页","section":"第一章","topics":["业务流程深水区"],"summary":"第一章章节页，引出大模型落地进入真实业务流程后的结构性难题。","sourceRef":"report:p7","fullTextPublished":false},
    {"page":8,"kind":"正文","section":"第一章","topics":["客户落地缺口","复合角色","传统分工失效"],"summary":"分析客户从 Demo 转向 Agent、系统接入、流程和指标后，传统售前、实施与产品分工出现的能力缺口。","sourceRef":"report:p8","fullTextPublished":false},
    {"page":9,"kind":"正文","section":"第一章","topics":["可运行 Demo","80/95/99","风险发现"],"summary":"解释 AI 项目为何需要尽早用真实问题和数据做可运行验证，并区分原型覆盖与生产精度。","sourceRef":"report:p9","fullTextPublished":false},
    {"page":10,"kind":"正文","section":"第一章","topics":["生产可靠性","长尾风险","审计回滚"],"summary":"说明从 95% 到 99% 需要权限、审计、回滚、人工接管和长尾评测，FDE 的价值集中在这一阶段。","sourceRef":"report:p10","fullTextPublished":false},
    {"page":11,"kind":"数据页","section":"第一章","topics":["招聘增长","组织路径","商业效果"],"summary":"汇总 2025 至 2026 年海外 FDE 招聘增长信号，并比较内部团队、生态放大和独立实体三种组织路径。","sourceRef":"report:p11","fullTextPublished":false},
    {"page":12,"kind":"正文","section":"第一章","topics":["中国市场","本地化难题","交付经济性"],"summary":"分析中国企业流程和数据基础的特点，指出国内 FDE 必须依靠 AI 工具、Skill、连接器和伙伴提高复用率。","sourceRef":"report:p12","fullTextPublished":false},
    {"page":13,"kind":"章节页","section":"第二章","topics":["前线学习机制"],"summary":"第二章章节页，提出 FDE 的本质是连接现场交付和组织学习的机制。","sourceRef":"report:p13","fullTextPublished":false},
    {"page":14,"kind":"正文","section":"第二章","topics":["重交付矛盾","规模经济","Echo 与 Delta"],"summary":"说明重交付容易依赖个人和堆人，FDE 只有把现场经验转成平台能力，才能摆脱人力项目模式。","sourceRef":"report:p14","fullTextPublished":false},
    {"page":15,"kind":"正文","section":"第二章","topics":["Delta 执行","Echo 抽象","双向反馈"],"summary":"定义 Delta 的原型、集成和调优职责，以及 Echo 的共性抽象和知识沉淀职责，强调二者协同。","sourceRef":"report:p15","fullTextPublished":false},
    {"page":16,"kind":"正文","section":"第二章","topics":["延迟收益","复用经济性","本体投资"],"summary":"解释知识和本体沉淀的成本先发生、收益后体现，需通过多个同类项目形成规模效应。","sourceRef":"report:p16","fullTextPublished":false},
    {"page":17,"kind":"正文","section":"第二章","topics":["知识载体演进","工作流","Multi-Agent","Skill"],"summary":"梳理知识图谱、本体、工作流、多 Agent 和 Skill 等沉淀载体的演进及维护成本差异。","sourceRef":"report:p17","fullTextPublished":false},
    {"page":18,"kind":"章节页","section":"第三章","topics":["Palantir 实践"],"summary":"第三章章节页，引出 Palantir 从 Gotham 到 AIP 的 FDE 与本体实践。","sourceRef":"report:p18","fullTextPublished":false},
    {"page":19,"kind":"案例页","section":"第三章","topics":["Palantir","高复杂客户","平台反馈"],"summary":"分析 Palantir 如何在高复杂客户现场解决问题，并将流程、权限和数据关系抽象为平台能力。","sourceRef":"report:p19","fullTextPublished":false},
    {"page":20,"kind":"正文","section":"第三章","topics":["Ontology","业务对象","语义统一"],"summary":"解释 Ontology 从按客户建表转向统一业务对象、关系和动作，降低跨客户重复建模成本。","sourceRef":"report:p20","fullTextPublished":false},
    {"page":21,"kind":"案例页","section":"第三章","topics":["AIP","Bootcamp","五天验证"],"summary":"说明 AIP 通过本体让模型理解业务语义，并以 Bootcamp 方式在客户现场快速验证高价值流程。","sourceRef":"report:p21","fullTextPublished":false},
    {"page":22,"kind":"正文","section":"第三章","topics":["成立条件","中国适配","反馈链"],"summary":"提出 Palantir 模式成立需要高质量反馈链、平台吸收能力和交付效率回流，并分析中国市场适配限制。","sourceRef":"report:p22","fullTextPublished":false},
    {"page":23,"kind":"章节页","section":"第四章","topics":["AI 成本变化","本体激活"],"summary":"第四章章节页，引出 AI 对 FDE 成本结构和本体使用门槛的改变。","sourceRef":"report:p23","fullTextPublished":false},
    {"page":24,"kind":"正文","section":"第四章","topics":["成本结构","原型成本","知识抽象"],"summary":"分析 AI 编程、自然语言交互和模型能力如何降低原型构建、需求表达与知识抽象成本。","sourceRef":"report:p24","fullTextPublished":false},
    {"page":25,"kind":"正文","section":"第四章","topics":["Skill","连接器","本体层","第二用户"],"summary":"区分 Skill、连接器和本体层的职责，并强调平台工具也要面向 FDE 的现场开发与反馈需要。","sourceRef":"report:p25","fullTextPublished":false},
    {"page":26,"kind":"正文","section":"第四章","topics":["AI Native","流程重构","局部提效"],"summary":"区分在旧流程中局部加 AI 与围绕 AI 重构流程、岗位和责任边界的 AI Native 路径。","sourceRef":"report:p26","fullTextPublished":false},
    {"page":27,"kind":"章节页","section":"第五章","topics":["Demo 到生产"],"summary":"第五章章节页，引出从演示系统进入生产环境的一线方法。","sourceRef":"report:p27","fullTextPublished":false},
    {"page":28,"kind":"正文","section":"第五章","topics":["生产差距","FDE 与 FDPM","真实环境"],"summary":"说明 Demo 进入真实业务后会暴露数据、权限、接口和组织问题，并提出 FDE 与 FDPM 配对。","sourceRef":"report:p28","fullTextPublished":false},
    {"page":29,"kind":"正文","section":"第五章","topics":["合规","业务指标","上线责任"],"summary":"强调金融、医疗和政务等场景的审计与接管要求，并要求把技术结果翻译为业务指标。","sourceRef":"report:p29","fullTextPublished":false},
    {"page":30,"kind":"框架页","section":"第五章","topics":["成熟度模型","阶段门槛","典型周期"],"summary":"给出从 Demo、试点到生产和扩展的成熟度判断框架，明确各阶段标志、任务和退出条件。","sourceRef":"report:p30","fullTextPublished":false},
    {"page":31,"kind":"章节页","section":"第六章","topics":["腾讯云实践"],"summary":"第六章章节页，引出腾讯云在平台、工具、行业团队和生态伙伴上的 FDE 实践。","sourceRef":"report:p31","fullTextPublished":false},
    {"page":32,"kind":"案例页","section":"第六章","topics":["云厂商场景","平台工具伙伴","轻重交付"],"summary":"分析国内云厂商面对的行业与区域差异，以及平台、工具和伙伴共同支撑交付的必要性。","sourceRef":"report:p32","fullTextPublished":false},
    {"page":33,"kind":"案例页","section":"第六章","topics":["教育行业","工作营","产品孵化"],"summary":"以教育行业说明如何通过工作营陪跑识别场景、搭建应用，并把高频需求沉淀为行业产品。","sourceRef":"report:p33","fullTextPublished":false},
    {"page":34,"kind":"案例页","section":"第六章","topics":["客户成功","陪跑师","高容错行业"],"summary":"描述客户成功团队承担场景识别、轻量搭建和持续运营的综合角色及适用行业。","sourceRef":"report:p34","fullTextPublished":false},
    {"page":35,"kind":"案例页","section":"第六章","topics":["CodeBuddy","WorkBuddy","全链路 FDE"],"summary":"描述 CodeBuddy 与 WorkBuddy 客户成功团队从高层认知、工程执行到商业闭环的全链路实践。","sourceRef":"report:p35","fullTextPublished":false},
    {"page":36,"kind":"框架页","section":"第六章","topics":["平台赋能","工具赋能","伙伴赋能"],"summary":"将云厂商 FDE 方法归纳为平台、工具和伙伴三层，并列出各层职责与关键产出。","sourceRef":"report:p36","fullTextPublished":false},
    {"page":37,"kind":"章节页","section":"第七章","topics":["落地困难","解决方法"],"summary":"第七章章节页，引出 FDE 在临时需求、强人依赖、利润和知识分散方面的现实困难。","sourceRef":"report:p37","fullTextPublished":false},
    {"page":38,"kind":"正文","section":"第七章","topics":["外包化风险","组织能力","共性沉淀"],"summary":"分析项目越做越重、知识越散和利润越薄的外包化风险，并强调结构化反馈与资产收尾。","sourceRef":"report:p38","fullTextPublished":false},
    {"page":39,"kind":"正文","section":"第七章","topics":["Land and Expand","续费扩展","采用运营"],"summary":"说明首个场景落地容易、持续扩展困难，要求在 Land 阶段设计使用、续费和场景扩展路径。","sourceRef":"report:p39","fullTextPublished":false},
    {"page":40,"kind":"框架页","section":"第七章","topics":["场景选择","重 FDE","经济性"],"summary":"给出判断场景是否适合重 FDE 的矩阵，综合业务关键性、复杂度、复用空间和合同价值。","sourceRef":"report:p40","fullTextPublished":false},
    {"page":41,"kind":"章节页","section":"第八章","topics":["人才","组织","考核"],"summary":"第八章章节页，引出 FDE 复合人才、组织归属、培养、晋升和激励问题。","sourceRef":"report:p41","fullTextPublished":false},
    {"page":42,"kind":"正文","section":"第八章","topics":["复合能力","人才短缺","岗位增长"],"summary":"分析既懂业务又能工程交付的人才短缺，并将招聘增长与复合能力缺口联系起来。","sourceRef":"report:p42","fullTextPublished":false},
    {"page":43,"kind":"正文","section":"第八章","topics":["内部培养","角色转型","晋升矛盾"],"summary":"主张从架构师、解决方案、交付和行业专家中培养 FDE，并讨论一线岗位的晋升与组织归属问题。","sourceRef":"report:p43","fullTextPublished":false},
    {"page":44,"kind":"框架页","section":"第八章","topics":["认证体系","项目认证","能力标准"],"summary":"给出准入、项目和高级认证框架，用真实交付、客户沟通、技术能力和资产沉淀作为通过标准。","sourceRef":"report:p44","fullTextPublished":false},
    {"page":45,"kind":"正文","section":"第八章","topics":["双人协作","共同指标","HR 建议"],"summary":"强调业务与技术角色需要共同驻场并承担同一业务指标，同时给出人才转型和考核建议。","sourceRef":"report:p45","fullTextPublished":false},
    {"page":46,"kind":"章节页","section":"第九章","topics":["全流程操作手册"],"summary":"第九章章节页，引出 Echo 与 Delta 从冷启动到规模复用的完整操作案例。","sourceRef":"report:p46","fullTextPublished":false},
    {"page":47,"kind":"案例页","section":"第九章","topics":["ERP 案例","角色配置","冷启动"],"summary":"设定虚构 ERP 行业案例，说明 Echo、Delta 和管理层在本体冷启动阶段的职责与产出。","sourceRef":"report:p47","fullTextPublished":false},
    {"page":48,"kind":"操作页","section":"第九章","topics":["本体评审","Delta 准备","管理层责任"],"summary":"列出 Delta 在首个客户前参与评审、学习本体和标记缺口的动作，以及管理层的资源保障。","sourceRef":"report:p48","fullTextPublished":false},
    {"page":49,"kind":"数据页","section":"第九章","topics":["定制工时","本体复用","内部定价"],"summary":"比较从零定制与复用本体的工时和内部成本，用于说明复用投资与项目经济性。","sourceRef":"report:p49","fullTextPublished":false},
    {"page":50,"kind":"操作页","section":"第九章","topics":["反馈单","费用抵扣","Echo 评审"],"summary":"说明 Delta 提交本体反馈、Echo 评估通用价值并通过费用抵扣激励高质量反馈的流程。","sourceRef":"report:p50","fullTextPublished":false},
    {"page":51,"kind":"操作页","section":"第九章","topics":["项目复盘","内部收入","早期飞轮"],"summary":"要求管理层在复盘中确认反馈采纳与结算结果，并进入第二到第四个客户的早期复用阶段。","sourceRef":"report:p51","fullTextPublished":false},
    {"page":52,"kind":"操作页","section":"第九章","topics":["版本节奏","核心路径","反馈分级"],"summary":"规定 Echo 的月度评审与版本发布节奏，并按通用性和价值对 Delta 反馈分级。","sourceRef":"report:p52","fullTextPublished":false},
    {"page":53,"kind":"操作页","section":"第九章","topics":["维护模式","跨行业复制","SOP"],"summary":"描述本体成熟后进入维护模式，Echo 将建设经验写成 SOP，并转向新行业。","sourceRef":"report:p53","fullTextPublished":false},
    {"page":54,"kind":"数据页","section":"第九章","topics":["成熟度指标","自助交付","本体组成"],"summary":"给出工时、配置覆盖和伙伴独立交付等成熟度指标，并列举成熟行业本体的组件构成。","sourceRef":"report:p54","fullTextPublished":false},
    {"page":55,"kind":"机制页","section":"第九章","topics":["定价","下单","抵扣","结算"],"summary":"定义本体内部结算的定价、使用、反馈抵扣和季度结算流程及责任人。","sourceRef":"report:p55","fullTextPublished":false},
    {"page":56,"kind":"方法页","section":"第九章","topics":["案例驱动","技术路径","知识抽象"],"summary":"补充案例驱动和结构化抽象两类本体沉淀路径，说明从业务流构建可复用语义资产的方法。","sourceRef":"report:p56","fullTextPublished":false},
    {"page":57,"kind":"章节页","section":"第十章","topics":["未来展望"],"summary":"第十章章节页，引出 Agent 能力增强后 FDE 角色、生态和度量方式的变化。","sourceRef":"report:p57","fullTextPublished":false},
    {"page":58,"kind":"正文","section":"第十章","topics":["角色分化","自动化","人类判断"],"summary":"判断技术执行会被 Agent 压缩，而问题定义、客户信任、行业知识和生态赋能的重要性会上升。","sourceRef":"report:p58","fullTextPublished":false},
    {"page":59,"kind":"趋势页","section":"第十章","topics":["生态化","度量化","区域化"],"summary":"提出原厂跑通标杆、伙伴规模复制，并用采用、任务、人效、成本、续费和扩展指标证明价值。","sourceRef":"report:p59","fullTextPublished":false},
    {"page":60,"kind":"机制页","section":"第十章","topics":["伙伴共担","资源调度","统一入口"],"summary":"建议建立 POC 风险共担、行业模板、联合支持和统一资源调度机制，改善伙伴交付经济性。","sourceRef":"report:p60","fullTextPublished":false},
    {"page":61,"kind":"附录","section":"研究方法","topics":["研究边界","岗位命名","外推限制"],"summary":"说明 FDE 定义和岗位名称仍在变化，并交代公开资料、访谈、招聘样本与案例的使用边界。","sourceRef":"report:p61","fullTextPublished":false},
    {"page":62,"kind":"附录","section":"数据口径","topics":["时间窗口","样本量","用途"],"summary":"列出海外招聘、国内招聘、公开材料和访谈的数据来源、采集窗口、样本规模及分析用途。","sourceRef":"report:p62","fullTextPublished":false},
    {"page":63,"kind":"附录","section":"术语表","topics":["FDE","FDPM","Echo","Delta","Ontology"],"summary":"定义报告使用的关键角色、机制和技术术语，减少岗位名称和概念口径差异。","sourceRef":"report:p63","fullTextPublished":false},
    {"page":64,"kind":"招聘附件","section":"执行摘要","topics":["海外样本","国内样本","薪酬概览"],"summary":"概览国内外招聘平台样本、海外有效岗位和薪资统计，以及国内显性 FDE 岗位与头部薪酬样本。","sourceRef":"report:p64","fullTextPublished":false},
    {"page":65,"kind":"招聘附件","section":"技能分析","topics":["技术技能","软技能","出现频率"],"summary":"统计编程、数据、云平台、架构等技术技能，以及客户沟通、问题拆解和跨职能推进等软技能。","sourceRef":"report:p65","fullTextPublished":false},
    {"page":66,"kind":"招聘附件","section":"学历经验","topics":["学历分布","经验层级","客户交付经历"],"summary":"分析海外岗位的学历和经验要求，指出 3 至 5 年是主流区间，客户交付经历被高频强调。","sourceRef":"report:p66","fullTextPublished":false},
    {"page":67,"kind":"招聘附件","section":"关键词与薪资","topics":["JD 关键词","出差","远程","海外薪资"],"summary":"列出岗位、技术和软技能高频词，并给出远程、出差比例及海外披露薪资的整体分布。","sourceRef":"report:p67","fullTextPublished":false},
    {"page":68,"kind":"招聘附件","section":"海外薪资","topics":["薪资分层","公司类型","影响因素"],"summary":"比较不同薪资区间与公司类型，分析赛道、技术稀缺性、地点和披露率对薪资统计的影响。","sourceRef":"report:p68","fullTextPublished":false},
    {"page":69,"kind":"招聘附件","section":"海外职责","topics":["部署集成","需求设计","客户成功","产品反馈"],"summary":"将海外 FDE 职责归纳为现场部署、系统集成、需求转化、客户协作和产品反馈等模块。","sourceRef":"report:p69","fullTextPublished":false},
    {"page":70,"kind":"招聘附件","section":"企业类型","topics":["AI 原生","国防政府","咨询","金融科技"],"summary":"比较 AI 原生、国防政府、咨询和金融科技企业的技术侧重、薪资、招聘偏好与工作方式。","sourceRef":"report:p70","fullTextPublished":false},
    {"page":71,"kind":"招聘附件","section":"企业类型","topics":["工业制造","地域分布","工作环境"],"summary":"补充工业制造类 FDE 的技术栈与薪资，并讨论不同地区和行业现场对岗位要求的影响。","sourceRef":"report:p71","fullTextPublished":false},
    {"page":72,"kind":"招聘附件","section":"海外趋势","topics":["薪资上行","级别分化","岗位增长"],"summary":"汇总外部薪酬报告和招聘趋势，比较中级、高级与顶级岗位的基本薪资及增长信号。","sourceRef":"report:p72","fullTextPublished":false},
    {"page":73,"kind":"招聘附件","section":"国内市场","topics":["数据来源","岗位增长","头部企业"],"summary":"列出海外数据来源，并概览 2026 年 6 至 7 月国内显性 FDE 招聘及头部企业岗位样本。","sourceRef":"report:p73","fullTextPublished":false},
    {"page":74,"kind":"招聘附件","section":"国内薪酬","topics":["企业样本","薪酬四层","经验要求"],"summary":"列举国内企业公开岗位样本，并将薪酬分为初级、中级、高级和负责人四个层级。","sourceRef":"report:p74","fullTextPublished":false},
    {"page":75,"kind":"招聘附件","section":"国内岗位","topics":["技术要求","业务要求","行业实践","政策信号"],"summary":"归纳国内 FDE 的编程、全栈、模型、Agent、客户和产品反馈要求，并列举企业与政策实践。","sourceRef":"report:p75","fullTextPublished":false},
    {"page":76,"kind":"招聘附件","section":"市场趋势","topics":["名称显性化","主体扩散","薪酬分层","中美差异"],"summary":"总结国内近半年岗位名称、雇主、薪酬和岗位内涵变化，并比较中美市场结构与沉淀能力。","sourceRef":"report:p76","fullTextPublished":false},
    {"page":77,"kind":"招聘附件","section":"雇主案例","topics":["Deloitte","Palantir","OpenAI"],"summary":"对比 Deloitte、Palantir 和 OpenAI 的薪资、地点、职责、经验要求、出差与成功标准。","sourceRef":"report:p77","fullTextPublished":false},
    {"page":78,"kind":"招聘附件","section":"雇主案例","topics":["Dynamo AI","Cognition AI","Tesla"],"summary":"对比 Dynamo AI、Cognition AI 和 Tesla 在治理、开发工具、制造现场及客户责任上的岗位差异。","sourceRef":"report:p78","fullTextPublished":false},
    {"page":79,"kind":"招聘附件","section":"雇主案例","topics":["Anthropic","跨企业比较","Agent 交付物"],"summary":"描述 Anthropic FDE 的 Agent 基础设施交付，并汇总企业间薪资、技术、出差和经验共性。","sourceRef":"report:p79","fullTextPublished":false},
    {"page":80,"kind":"参考文献","section":"参考文献","topics":["企业材料","行业报告","数据平台"],"summary":"列出 Palantir、OpenAI、Anthropic 等企业材料以及 FDE 招聘与薪酬研究来源。","sourceRef":"report:p80","fullTextPublished":false},
    {"page":81,"kind":"参考文献","section":"参考文献","topics":["访谈","招聘平台","媒体","腾讯实践"],"summary":"补充 IDC、LinkedIn、公开访谈、招聘平台、媒体报道、腾讯研究与产品材料等来源。","sourceRef":"report:p81","fullTextPublished":false},
    {"page":82,"kind":"结束页","section":"报告结语","topics":["工程方法","深入业务"],"summary":"结束页强调大模型不会自动进入生产，需要深入业务现场的新工程方法。","sourceRef":"report:p82","fullTextPublished":false},
    {"page":83,"kind":"空白页","section":"文档结束","topics":["版式留白"],"summary":"PDF 最后一页为空白版式页，不包含新的报告正文或数据。","sourceRef":"report:p83","fullTextPublished":false}
  ]
}