[{"data":1,"prerenderedAt":76},["ShallowReactive",2],{"news-list-all":3,"news-counts":72},[4,19,33,46,59],{"slug":5,"title":6,"summary":9,"body":12,"category":15,"date":16,"featured":17,"published":17,"readMinutes":18},"fde-forward-deployed-engineer",{"zh":7,"en":8},"FDE 前沿部署工程师：AI 落地最后一公里靠谁","FDE — Who Delivers AI Into Production?",{"zh":10,"en":11},"大模型能力很强，但 95% 的企业生成式 AI 项目拿不到可衡量回报。FDE（Forward Deployed Engineer）是驻场把 AI 推进生产环境、对业务结果负责的工程师角色。","LLMs are capable, yet ~95% of enterprise generative-AI programs fail to deliver measurable returns. A Forward Deployed Engineer (FDE) is the engineer who takes AI into production on site and owns the business outcome.",{"zh":13,"en":14},"## 模型不是瓶颈，落地才是\n\n企业在生成式 AI 上的投入持续增长，但真正跑进生产环境并产生可衡量回报的比例并不高。MIT 的企业 AI 使用调查报告显示，约 95% 的组织未获得可衡量的投资回报；另一份 2025 年企业 AI 调研中，96% 的高管计划加码生成式 AI 投入，却只有 36% 成功部署到生产环境。\n\n原因不在模型。GPT、Claude、Qwen 的能力已经足够，真正的堵点集中在企业\"最后一公里\"：\n\n- 数据散落在 CRM、ERP、邮件、Excel 与各部门，口径不一致\n- 权限没人敢开、审计没人敢签、法务只报风险\n- IT 怕出事故，业务部门怕被替代，跨部门无人拍板\n- Demo 能跑，一进生产就死\n\n## FDE 是什么\n\n**FDE（Forward Deployed Engineer，前沿部署工程师）** 是驻场把通用大模型能力部署进企业真实生产环境、并对最终业务结果负责的工程师。\n\n三个词拆开看更清楚：\n\n- **Forward（前沿）**：不在总部。FDE 不在产品团队的聊天窗口里异步回应，他在客户会议室里现场。\n- **Deployed（已部署）**：已经派驻到客户那边，在客户的工单系统、脏数据、旧权限里干活。\n- **Engineer（工程师）**：交付的是生产代码，不是 PPT；他写的代码要在客户面前真能跑、真被使用。\n\n这一模式源自 Palantir 的 Delta 模式。2003 年 Palantir 试图把数据分析软件卖给 CIA 等机构，客户提不出明确需求，创始人干脆把工程师直接派到客户现场驻场开发，边听边改，最终长出了这套方法论。2020 年上市后被硅谷广泛关注，2024–2026 年 AI 大爆发后，OpenAI、Anthropic 等纷纷复制，FDE 从一家公司的内部角色演变为全行业配置。\n\n## 五步工作流\n\n1. **驻场理解业务**：与业务、IT 一线同室办公，搞清真实流程、卡点与决策链。\n2. **定义真问题**：把业务痛点翻译成可度量的技术问题，明确 AI 要影响的是哪个业务指标。\n3. **快速搭建原型**：用最小可用路径打通数据与系统接口，尽快拿出可运行的原型。\n4. **现场迭代验证**：在真实环境与真实数据上反复验证，把响应速度、准确率与业务指标对齐。\n5. **沉淀可复用能力**：把现场经验抽象为行业模板、组件与规范，降低下一个客户的落地周期与成本。\n\n## FDE 和传统交付有什么不同\n\n| 对比维度 | 传统外包 \u002F 售前 \u002F 架构师 | FDE |\n| --- | --- | --- |\n| 工作方式 | 远程交付、按需求单逐项完成任务 | 驻场常驻，与业务同频迭代 |\n| 交付物 | 文档、Demo 与方案 PPT | 生产环境可运行、有人真正使用的系统 |\n| 对结果负责 | 对工作量与验收单负责 | 对业务指标负责：工时、成本、响应速度 |\n| 数据现实 | 假设数据是干净的 | 把大部分精力花在清洗散落的历史数据上 |\n| 经验回流 | 项目结束即结束 | 现场经验回流产品与行业模板，缩短下一个项目周期 |\n\n## 企业该怎么用 FDE\n\nFDE 模式并不等于永久依赖。常见的正确姿势是**先借力、再内化**：先用 FDE 快速验证并兑现价值，同时让自有 IT 团队跟着参与，把能力逐步接回自己手中。\n\n以下情况更适合引入 FDE：\n\n- 已有明确的 AI 场景，但缺能把 AI 推进生产环境的人\n- 数据分散、口径不清，需要有人驻场治理\n- 需要跨部门推动，需要一个既懂技术又能让各方拍板的权威角色\n- 想先小成本验证效果，再决定是否自建团队\n\n## 企数管家的 FDE\n\n在企数管家的服务体系中，FDE 是「数字化管家服务」里 AI 落地的核心交付形态，对应 PPDIOO 六阶段中的实施（Implementation）环节。\n\n它不是孤立的一个角色，而是与三位专属管家协同：BAM 客户经理管家对齐业务目标与验收标准，TAM 技术经理管家保障系统架构、集成与运维稳定性，AIM 智能 AI 管家提供模型与数据能力支撑并持续监控运行状态。项目结束后，现场经验会被沉淀为行业模板与组件，回流自研 AITbuddy 平台，让下一个客户的落地周期更短、成本更低。\n\n如需了解 FDE 在您所在行业的具体落地方式，欢迎通过[联系我们](\u002Fcontact)预约落地诊断。","## The model is not the bottleneck — deployment is\n\nEnterprise spending on generative AI keeps rising, yet the share of programs that actually reach production with measurable returns remains low. MIT's enterprise AI usage research reports that roughly 95% of organizations see no measurable return; a separate 2025 enterprise AI survey found 96% of executives plan to increase generative-AI investment, but only 36% have deployed it to production.\n\nThe model is not the problem. GPT, Claude and Qwen are capable enough. The real blockage sits in the \"last mile\":\n\n- Data is scattered across CRM, ERP, email, Excel and departments, with inconsistent definitions\n- Nobody dares open permissions or sign off on audits; legal only reports risk\n- IT fears incidents, business units fear replacement, and nobody owns cross-department decisions\n- Demos run fine and die in production\n\n## What is an FDE\n\nA **Forward Deployed Engineer (FDE)** is an engineer stationed on site who deploys general-purpose LLM capabilities into the client's real production environment and stays accountable for the business outcome.\n\nThe three words matter:\n\n- **Forward** — not at headquarters. An FDE does not respond asynchronously from a product chat room; they are in the client's meeting room.\n- **Deployed** — already embedded with the client, working inside their ticket system, dirty data and legacy permissions.\n- **Engineer** — ships production code, not slides; the code must actually run, in front of the client, and actually be used.\n\nThe model originates from Palantir's Delta approach. In 2003 Palantir could not get crisp requirements from intelligence agencies, so founders simply stationed engineers with clients, listening and iterating — and the methodology grew from there. It went public in 2020 and drew wide attention in Silicon Valley; after the 2024–2026 AI boom, OpenAI, Anthropic and others copied it, and the FDE evolved from one company's internal role into an industry-wide standard.\n\n## The five-step workflow\n\n1. **Understand the business on site** — work alongside business and IT to map real processes, blockers and decision chains.\n2. **Define the real problem** — translate business pain into measurable technical problems: which metric must AI move?\n3. **Build a fast prototype** — connect data and system interfaces via the shortest viable path.\n4. **Iterate and validate on site** — validate against real data until latency, accuracy and business metrics align.\n5. **Codify reusable capability** — abstract field experience into templates, components and standards.\n\n## How FDE differs from traditional delivery\n\n| Dimension | Outsourcing \u002F presales \u002F architect | FDE |\n| --- | --- | --- |\n| Working mode | Remote delivery against a requirements list | Embedded on site, iterating at business speed |\n| Deliverable | Documents, demos and solution decks | A running system in production that people use |\n| Accountability | Effort and acceptance sheets | Business metrics: hours, cost, response time |\n| Data reality | Assumes clean data | Spends most effort cleansing fragmented legacy data |\n| Knowledge feedback | Ends at project close | Field experience flows back into product templates |\n\n## How enterprises should use FDEs\n\nFDE does not mean permanent dependency. The common pattern is **borrow first, internalize later**: use FDEs to validate quickly and realize value, while your own IT team participates and gradually takes the capability in-house.\n\nFDEs fit best when:\n\n- You have a clear AI scenario but no one who can push it into production\n- Data is fragmented with unclear definitions and needs on-site governance\n- Cross-department decisions need a credible authority who understands both technology and stakeholders\n- You want to validate value at modest cost before deciding to build a team\n\n## FDE at Enterprise Digital Steward\n\nWithin our service system, the FDE is the core delivery model for AI adoption inside our Managed Digital Steward service, corresponding to the Implementation stage of the PPDIOO framework.\n\nIt is not a standalone role — it runs with three dedicated stewards: the BAM Client Manager Steward aligns business goals and acceptance criteria, the TAM Technical Manager Steward secures architecture, integration and operational stability, and the AIM Smart AI Steward provides model and data capability with continuous monitoring. After each project, field experience is codified into industry templates and components and fed back into our in-house AITbuddy platform, so the next client deploys faster and at lower cost.\n\nTo see how FDE applies to your industry, [contact us](\u002Fcontact) to book a deployment diagnostic.","insight","2026-10-11",true,6,{"slug":20,"title":21,"summary":24,"body":27,"category":30,"date":31,"published":17,"readMinutes":32},"edsteward-website-launch",{"zh":22,"en":23},"企数管家官网正式上线，数字化服务入口全面开放","Enterprise Digital Steward Website Goes Live",{"zh":25,"en":26},"官网提供四项核心服务介绍与在线留资入口，选型询价可直接跳转 aitbuddy.cn 完成。","The website presents four core services with online lead capture; selection and pricing flow directly into aitbuddy.cn.",{"zh":28,"en":29},"## 官网正式上线\n\n企数管家官网（edsteward.cn）已正式上线，为企业提供四项核心服务的完整介绍与在线留资入口。\n\n**网站提供的内容：**\n\n- **数字化管家服务**：托管式 IT \u002F 数字化管家，定期巡检、主动维护\n- **AIT 产品选型·询价**：多品牌透明选型，帮企业在预算内选到合适方案\n- **数字化转型咨询（DTMM）**：基于 GB\u002FT 43439-2023 的成熟度评估与路径规划\n- **IT 资产·运维管理**：资产台账、一机一码 SN 池、运维记录全流程留痕\n\n**配套工具：**\n\n我们同步上线了[免费数字化成熟度自测](\u002Fself-check)，回答 6 个问题即可了解企业数字化现状与改进方向。\n\n**服务衔接：**\n\n官网的选型、咨询与资产服务入口均直接对接业务平台 aitbuddy.cn，参数自动携带来源标记，便于后续对接与服务跟踪。\n\n欢迎访问 [edsteward.cn](https:\u002F\u002Fedsteward.cn) 了解详情，或直接[联系我们](\u002Fcontact)。","## Website now live\n\nThe Enterprise Digital Steward website (edsteward.cn) is now live, presenting four core services with an online lead capture path.\n\n**What the site offers:**\n\n- **Managed Digital Steward service**: outsourced IT and digital stewardship with regular inspection and proactive maintenance\n- **AIT product selection and pricing**: multi-brand transparent selection within your budget\n- **Digital transformation consulting (DTMM)**: maturity assessment and roadmap based on GB\u002FT 43439-2023\n- **IT asset and operations management**: asset ledger, per-unit SN tracking, and full maintenance records\n\n**Companion tool:**\n\nWe also launched a [Free Digital Maturity Self-Check](\u002Fself-check) — answer 6 questions to learn your current state and improvement direction.\n\n**Service continuity:**\n\nSelection, consulting, and asset service entries link directly into aitbuddy.cn, with source parameters carried automatically for follow-up and tracking.\n\nVisit [edsteward.cn](https:\u002F\u002Fedsteward.cn) or [contact us](\u002Fcontact).","company","2026-10-09",2,{"slug":34,"title":35,"summary":38,"body":41,"category":15,"date":44,"featured":17,"published":17,"readMinutes":45},"what-is-ait-digital-steward",{"zh":36,"en":37},"什么是 AIT 数字化管家？与传统 IT 外包有什么不同","What is an AIT Digital Steward? How It Differs from Traditional IT Outsourcing",{"zh":39,"en":40},"AIT 把人工智能与信息技术融合，不是简单外包 IT 人力，而是用三人专属管家机制对数字化全生命周期负责。","AIT fuses AI with IT — not simply outsourcing IT staff, but making three dedicated stewards accountable for the full digitalization lifecycle.",{"zh":42,"en":43},"## 传统 IT 外包的边界\n\n传统 IT 外包解决的是\"人手问题\"：把系统运维、故障处理、设备管理按人天或人月外包给服务商。这类合作通常有三个特点：\n\n- **按工时计费**，服务范围以人天为边界\n- **按需响应**，问题出现才介入\n- **责任止于系统可用**，不关注业务价值是否实现\n\n## AIT 数字化的不同\n\nAIT（AI + IT）数字化管家服务在此基础上做了三件事：\n\n**第一，服务对象从\"系统\"变为\"业务目标\"。** 我们不只保证系统可用，而是承接明确的业务目标（例如\"把资产盘点周期从 15 天压缩到 3 天\"）。\n\n**第二，三人专属机制替代按需响应。** BAM 客户经理管家负责业务目标与方案，TAM 技术经理管家负责技术交付与稳定性，AIM 智能 AI 管家负责持续监控与优化。客户不需要自己判断该找谁。\n\n**第三，PPDIOO 六阶段保证全生命周期。** 从规划、准备、设计、实施、运营到优化，每个阶段都有明确的交付物与验收标准，避免上线即失管。\n\n## 适用场景\n\n- 中小微企业：没有专职 IT 团队，但业务已依赖信息系统\n- 集团分支机构：IT 能力由总部统一提供，但本地需要响应\n- 产业园区：需要为入驻企业提供统一的数字化服务底座\n\n如果您正在评估数字化服务方案，欢迎通过官网「联系我们」与我们沟通，或在 aitbuddy.cn 提交选型需求。","## The boundaries of traditional IT outsourcing\n\nTraditional IT outsourcing solves a \"manpower\" problem: system operations, incident handling, and device management outsourced by person-day or person-month. Such partnerships typically have three traits:\n\n- **Billed by hours**, with scope bounded by person-days\n- **Reactive**, stepping in only when issues appear\n- **Accountability ends at system availability**, without concern for business value\n\n## What makes AIT different\n\nAIT (AI + IT) digital stewardship builds on this in three ways:\n\n**First, the service object shifts from \"systems\" to \"business goals.\"** We don't just keep systems available — we take on explicit business objectives (for example, \"shorten the asset inventory cycle from 15 days to 3\").\n\n**Second, a three-person dedicated mechanism replaces reactive response.** A BAM Client Manager Steward owns business goals and solution design, a TAM Technical Manager Steward owns technical delivery and stability, and an AIM Smart AI Steward owns continuous monitoring and optimization. Clients never need to figure out whom to call.\n\n**Third, PPDIOO six stages ensure full-lifecycle coverage.** From planning, preparation, design, implementation, operations to optimization, each stage has defined deliverables and acceptance criteria — preventing the common \"go-live then neglect\" failure.\n\n## Where it fits\n\n- **SMEsm**: no dedicated IT team, yet the business already depends on information systems\n- **Group branches**: IT capability provided by headquarters, but local response is needed\n- **Industrial parks**: a shared digital service baseline for tenant enterprises\n\nIf you are evaluating a digital service partner, reach out via the website Contact page or submit a selection request on aitbuddy.cn.","2026-10-08",4,{"slug":47,"title":48,"summary":51,"body":54,"category":15,"date":57,"published":17,"readMinutes":58},"dtmm-maturity-model-explained",{"zh":49,"en":50},"DTMM 数字化成熟度模型：企业该在哪一级？","The DTMM Maturity Model: Which Level Is Your Enterprise?",{"zh":52,"en":53},"依据 GB\u002FT 43439-2023 及其衍生态，数字化成熟度分为 L1–L5 五个等级，并以数据治理为核心引导维度。了解自己在哪一级，才能知道下一步该补什么。","Under GB\u002FT 43439-2023 and its derivatives, digital maturity spans five levels (L1–L5), guided by data governance as the core dimension. Knowing your level tells you what to fix next.",{"zh":55,"en":56},"## 为什么需要成熟度评估\n\n数字化投入最常见的失败方式，是\"哪里痛就治哪里\"——设备坏了买设备，系统卡了上系统。成熟度评估的价值在于给出全局坐标，避免局部最优。\n\n## 五个等级\n\n**L1 · 初始级**：以手工与 Excel 为主，缺少系统或仅有纸笔化系统，数据与流程都在个人手里。\n\n**L2 · 基础级**：单点系统上线，例如 ERP 或 CRM 单独运行，跨部门仍靠人工传递数据。\n\n**L3 · 集成级**：系统打通、数据贯通，ERP + MES + WMS 完成集成，典型如智造工厂的\"数据一张网\"。\n\n**L4 · 优化级**：数据驱动与智能应用，建立数据中台与 BI，AI 深度嵌入关键业务场景。\n\n**L5 · 引领级**：AI 原生与自主进化，数字孪生、虚实融合大模型、自进化 Agent 等形态出现。\n\n## 评估维度\n\n成熟度统一从**数据治理**维度引导，并指向数据管理层：\n\n| 维度 | 关注点 |\n| --- | --- |\n| 数据治理（核心引导维度） | 数据资产的管理、标准、质量与权属 |\n| 系统建设 | IT 基础设施与业务系统的覆盖度 |\n| 系统集成 | 跨系统打通与数据贯通程度 |\n| 智能应用 | AI 与数据分析在业务中的实际渗透 |\n\n## 怎么用\n\n在我们的[免费数字化成熟度自测](\u002Fself-check)中，您只需回答 6 个问题（约 3 分钟），即可得到所在等级、分维度得分与针对性建议。测试完成后可预约专属顾问获取详细解读。","## Why maturity assessment matters\n\nThe most common failure mode of digital investment is treating symptoms: buy hardware when devices break, deploy systems when processes stall. A maturity assessment provides a global coordinate to avoid local optimization.\n\n## The five levels\n\n**L1 · Initial**: manual work and spreadsheets dominate; no system, or paper-based ones; data and processes live with individuals.\n\n**L2 · Basic**: point systems go live, such as a standalone ERP or CRM, with data still passed manually between departments.\n\n**L3 · Integrated**: systems are connected and data flows through — ERP + MES + WMS integrated, as in a smart factory running on a single data network.\n\n**L4 · Optimized**: data-driven and intelligent, with a data platform and BI in place and AI deeply embedded in key processes.\n\n**L5 · Leading**: AI-native and self-evolving — digital twins, digital-physical foundation models, self-evolving agents.\n\n## Assessment dimensions\n\nMaturity is guided uniformly by **data governance**, pointing toward the data management layer:\n\n| Dimension | Focus |\n| --- | --- |\n| Data governance (core guiding dimension) | Management, standards, quality and ownership of data assets |\n| System building | Coverage of IT infrastructure and business systems |\n| System integration | Degree of cross-system connectivity and data flow |\n| Intelligent application | Actual penetration of AI and analytics into the business |\n| Organization | Transformation mechanisms and cross-team collaboration |\n\n## How to use it\n\nIn our [Free Digital Maturity Self-Check](\u002Fself-check), you answer 6 questions (about 3 minutes) to get your level, per-dimension scores, and tailored advice. You can then book a dedicated advisor for detailed interpretation.","2026-10-06",3,{"slug":60,"title":61,"summary":64,"body":67,"category":70,"date":71,"published":17,"readMinutes":45},"sn-pool-asset-management",{"zh":62,"en":63},"一机一码（SN 池）：IT 资产管理的关键抓手","Per-Unit SN Tracking: The Key to IT Asset Management",{"zh":65,"en":66},"资产台账为什么总对不上？问题往往出在\"编号\"而不是\"盘点\"。一机一码是成本最低、见效最快的解法。","Why do asset ledgers keep mismatching? The problem is usually numbering, not stocktaking. Per-unit SN tracking is the cheapest fix.",{"zh":68,"en":69},"## 常见困境\n\n很多企业的 IT 资产台账不是没有，而是\"对不上\"：\n\n- 设备领用后未登记调拨去向\n- 同一台设备维修换件后编号变化\n- 盘点时靠人记忆核对，效率低\n- 设备报废\u002F闲置未及时下架，台账虚高\n\n根源在于**编号粒度不够**。很多台账记到\"型号\"级别（如\"ThinkPad X1 Carbon 2023 共 12 台\"），一旦混入不同批次\u002F配置，账实就再也核不清。\n\n## 一机一码怎么做\n\n**SN 池**：每台新设备出厂即生成唯一序列号，登记入池。\n\n1. **入库**：SN 登记，绑定型号、批次、到货日期\n2. **领用**：SN 绑定使用人与部门，领用记录留痕\n3. **调拨**：SN 跟随设备流转，全程可追溯\n4. **维修**：换件记录挂在同一 SN 下，不新建条目\n5. **盘点**：按 SN 扫码核对，一天能盘完几百台\n6. **下架**：报废\u002F闲置时 SN 状态置为已下架\n\n## 带来的变化\n\n- **盘点效率**：从\"翻台账对实物\"变成\"扫码核对\"，单人单日可盘 200+ 台\n- **账实一致**：每台设备的状态、位置、责任人一清二楚\n- **成本决策**：设备使用年限、故障率数据支撑预算与替换决策\n- **安全合规**：设备离场有据可查，避免资产流失\n\n## 落地建议\n\n如果企业资产量在 50 台以上，建议将\"一机一码\"作为 IT 资产管理的第一个抓手——它见效快、改造成本低，且是后续 AI 运维与预测性维护的数据基础。\n\n了解详情可访问[服务·产品](\u002Fservices)页，或直接[联系我们](\u002Fcontact)获取资产台账方案建议。","## A common predicament\n\nMany enterprises do have IT asset ledgers — they just never match reality:\n\n- Devices go out on assignment but the transfer is never recorded\n- Repairs that swap parts change the device identifier\n- Stocktaking relies on memory, which is slow and error-prone\n- Retired or idle devices are never delisted, inflating the ledger\n\nThe root cause is **insufficient granularity**. Many ledgers record at the model level (e.g., \"ThinkPad X1 Carbon 2023, 12 units\"). Once units from different batches or configurations mix, reconciliation becomes impossible.\n\n## How per-unit SN tracking works\n\n**An SN pool**: every new device gets a unique serial number at intake.\n\n1. **Intake**: register the SN, binding model, batch, and arrival date\n2. **Assignment**: bind the SN to a user and department, with a record\n3. **Transfer**: the SN follows the device, fully traceable\n4. **Repair**: part-replacement history attaches to the same SN — no new entries\n5. **Stocktake**: scan to reconcile — one person can finish hundreds of units a day\n6. **Delisting**: mark the SN as retired when a device is scrapped or idled\n\n## What changes\n\n- **Stocktaking efficiency**: from \"matching ledgers to physical assets\" to \"scanning to verify\" — one person, 200+ units per day\n- **Ledger accuracy**: every device's status, location, and owner is unambiguous\n- **Cost decisions**: usage age and failure-rate data underpin budgeting and replacement\n- **Security and compliance**: device departures are traceable, preventing asset leakage\n\n## Where to start\n\nIf you manage 50+ devices, per-unit SN tracking is the best first move — fast results, low retrofit cost, and it forms the data foundation for AI operations and predictive maintenance later.\n\nSee [Services](\u002Fservices) for details, or [Contact us](\u002Fcontact) for a tailored asset ledger plan.","tech","2026-10-05",{"all":73,"insight":58,"company":74,"tech":74,"media":75},5,1,0,1791962663997]