Chinese-Style FDE / 中国式 FDE

Definition

Chinese-style FDE is FDE practice under Chinese enterprise conditions, where AI deployment often begins before agent construction with business-access negotiation, data and document preparation, workflow reconstruction, permission design, and customer-side authority alignment.

Current Synthesis

The current synthesis treats Chinese-style FDE as a local operating pattern rather than a fashionable title. 瓴羊’s account frames the ideal as a business-effect team capability: AI acuity, industry depth, data breadth, and customer-side expert coaching turn agent projects into measurable service, sales, marketing, or operations outcomes. 申越’s field account supplies the less polished version: even a competent FDE can be stuck as a vendor-side rescuer if the customer cannot provide real business access, internal rules, usable documents, or a decision-maker who can mobilize departments.

This makes Chinese-style FDE a bridge between Business-Led AI Transformation and China Enterprise AI System Debt. Model quality matters, but the recurring constraint is whether the enterprise has enough process, data, authority, acceptance criteria, and human cooperation for probabilistic AI systems to perform real work.

Key Claims

• Chinese-style FDE starts from measurable business effects, but the first bottleneck may be gaining access to real business teams rather than selecting a model or building a demo.
• It often has to repair China Enterprise AI System Debt through workflow mapping, data governance, permission setup, knowledge-base cleanup, internal-rule extraction, and system integration.
• The role is usually a team or cross-organization capability, not a single heroic engineer: business analysts, AI architects, customer experts, and vendor-side communicators all appear in the evidence.
• Communication, industry knowledge, and AI judgment can be more decisive than narrow coding skill when customer authority, acceptance, and cooperation are unresolved.
• State-owned and private enterprises create different pressure patterns: state-owned projects may stall on ownership and business access, while private firms press harder on ROI, usage, token cost, and labor-saving claims.
• Production deployment remains bounded by probabilistic model behavior, human review, document quality, and whether AI is being asked to solve a genuine business workflow rather than a deeper demand problem.

Evidence

• Business-effect frame: E248｜一个“催发货”AI要跑通260步，和阿里瓴羊彭新宇聊聊中国式FDE defines China-side FDE around AI acuity, industry depth, data breadth, and delivery of business outcomes rather than functions.
• Team capability: E248｜一个“催发货”AI要跑通260步，和阿里瓴羊彭新宇聊聊中国式FDE describes a business analyst, AI architect, and customer-side expert-coach mix, while EP128 从 Palantir 到 OpenAI：FDE 会成为 AI 时代最重要的新岗位？ 🧬 uses Palantir’s Echo/Delta split to show that business interpretation and engineering delivery can be separate roles.
• Workflow triage and deployment rigor: E240｜OpenAI联手PE砸下40亿美元，聊聊硅谷最火新职位FDE grounds FDE in use-case selection, API validation, rollout monitoring, deterministic-versus-AI workflow decomposition, and human review.
• Enterprise-software substrate: 174: AI冲击企业软件巨头？与SAP原欣聊大模型to B的颠覆与边界 says enterprise AI must reconstruct business objects, ontology, trusted data, auditability, and process memory before agents can act reliably.
• Customer-authority bottleneck: 一个中国 FDE 的光环、落差与「救火」日常｜S10E27 shows a state-owned project turning only after a new responsible person could coordinate eight or nine business departments and let the team gather real materials and workflows.
• ROI and foundation work: 一个中国 FDE 的光环、落差与「救火」日常｜S10E27 describes private-company pressure around usage, token spend, headcount savings, and revenue, plus document cleanup and contract-review rule extraction as prerequisites for usable agents.

Counterevidence & Qualifications

The evidence base is interview-heavy and should not be treated as a representative survey of all Chinese enterprise AI projects. The 瓴羊 source describes a more structured vendor/team capability, while 申越’s episode emphasizes lower-power vendor-side firefighting; together they show a range of practice rather than a single settled operating model. Claims about ROI, usage, and headcount savings remain source-scoped unless later sources provide audited project metrics.

What Changed

• Migrated the page to synthesis-v1 while preserving the existing source order and adding the What’s Next S10E27 source once.
• Shifted the synthesis from foundation-building alone to foundation-building plus customer-side authority and business-access negotiation.
• Added state-owned versus private-enterprise pressure patterns from 申越’s field cases.
• Added explicit qualifications around vendor-side power, probabilistic model limits, and demand problems AI may not solve.

Related Concepts

• Forward Deployed Engineer - broader deployment role localized by this concept.
• Business-Led AI Transformation - organization-level transformation frame that Chinese-style FDE tries to operationalize.
• China Enterprise AI System Debt - data, workflow, and software-foundation gap this practice often has to repair.
• Enterprise Operational Memory - process and business-object memory agents need before they can act reliably.
• Enterprise Data Activation - data-to-action layer required for enterprise agents.
• AI Workflow Triage - workflow decomposition method used to decide where AI belongs.
• Enterprise AI ROI Audit - measurement pressure that makes FDE work accountable to business value.