Reconstructing a professional-services firm from public information把一家专业服务公司
从公开信息重构上引擎

Client & challenge客户与挑战

A global financial & risk advisory firm — three pillars (Financial Advisory / Risk Advisory / Business Services), 32 client-facing service lines, organized around client engagements rather than functional silos. The challenge: process frameworks are authored from the perspective of a company running its own operations. This firm sells those operations to others. Does the framework still describe it?一家全球金融与风险咨询公司——三大板块(财务咨询/风险咨询/业务服务)、32 条对外服务线，围绕客户组队而非职能孤岛。挑战在于：流程框架都是从“企业经营自己”的视角写的，而这家公司把这些经营职能卖给别人。框架还描述得了它吗？

Method · the engine is the trunk方法 · 引擎当主干

Every service line runs the same chain: semantic anchor → process blueprint → KPIs → responsibilities → cross-domain insight → provenance. The test is the hollow-run check: strip the ontology IRIs, the graph traversal and the provenance and the conclusions must collapse. Provenance is dual — every conclusion points both to a source public page and to a node IRI + derivation rule + confidence. Top-1 vector hits are never trusted without lineage confirmation (one 0.741 hit was confirmed a lexical false-friend and recorded as boundary).每条服务线走同一条链：语义锚定 → 流程蓝图 → KPI 挂接 → 职责派生 → 跨域洞察 → 全程溯源。判据是空转检验：抽掉本体 IRI、图遍历与溯源，结论应整个塌掉。全程双重溯源——每条结论既指向源公开网页，又指向节点 IRI + 派生规则 + 置信度。禁信 top-1 向量命中、必用血缘确认(一处 0.741 高分被确认是词面假朋友、判为 boundary)。

The engine-emergent finding: three anchors, one gradient引擎涌现的核心发现：三锚点与覆盖梯度

Graph traversal forces a structural conclusion: a professional-services firm sells, as its product, functions the framework models only as a client's internal activity — or not at all. So every line shows a 3-anchor pattern:图遍历逼出一个结构性结论：专业服务公司把作为产品出售的职能，在框架里往往只被建模为“客户的内部活动”、或根本没有。于是每条线呈现三锚点：

• Delivery spine — shared, strong (0.69–0.74). How any engagement is staffed, executed and closed; identical across all lines (assign resources, initiate, complete, provide service). The clean anchor.交付主轴——共用、强(0.69–0.74)。如何配置/执行/收尾一次 engagement，对全部服务线一致(资源指派、启动、收尾、提供服务)。这是干净的锚。
• Domain content — per-line, weak-to-absent. The expertise sold maps only to “what the client would otherwise do in-house” — an externalized internal function.领域内容——逐线、弱到缺。所卖专业内容只能映到“客户本会在内部做的那件事”——外部化的内部职能。
• Buyer's mirror. The framework sees the firm only from the client's side — as “engage an external advisor”. It never sees the supply side of professional services.买方镜像。框架只从“客户聘请外部顾问”这一侧看到这家公司，从不见专业服务的供给侧。

Across 32 lines the domain coverage computes to a gradient — 6 clean · 6 boundary · 13 weak · 7 gap — strictly tracking how far a service externalizes a support function the framework already models. Cyber, compliance and fixed-asset land cleanly (one even ships a catalog KPI measuring the cost of outsourced accounting — the framework already anticipated the outsourcing this firm sells); regulated administration, fairness opinions and expert testimony are gaps.32 条线的领域覆盖算出一条梯度——6 clean · 6 boundary · 13 weak · 7 gap——严格追随“这项服务在多大程度上外部化了框架已建模的支撑职能”。网络安全、合规、固定资产干净落位(后者的库内 KPI 甚至在度量外包会计的成本——框架早已预见这家公司在卖的外包)；受监管行政、公允性意见、专家证词则是缺口。

The supply-chain “extension” test供应链能力的“外延”检验

Choosing a services firm was deliberate — to test whether SCOR supply-chain analysis, built for physical products, extends to a “people + engagements” service supply chain. The answer is bounded partial support: the delivery spine plus the two existing service-SCOR variants cover ~60–70% (talent sourcing, staffing, fulfilment); what's missing — the service-side Source/Order/Return/Orchestrate variants, the regulated-administration nodes and the near-empty service-KPI layer — is written up as ontology-extension proposals in their own namespace, leaving the core untouched and reversible.选专业服务公司当案例是刻意的——检验为实物产品而建的 SCOR 供应链分析能否外延到“人+项目”的服务供应链。结论是有边界的部分支持：交付主轴 + 既有两个服务-SCOR 变体约覆盖 60–70%(人才寻源、配置、履行)；缺的服务侧 Source/Order/Return/Orchestrate 变体、受监管行政节点与几乎为空的服务 KPI 层，被写成自有命名空间的本体扩展提案——核心零改、可逆。

The deeper layer: neutrality is a position, not a disposition更深的一层：中立是位置，不是精神

The hardest cluster inside those gaps — regulated administration, fairness opinions, expert testimony — is not only missing nodes. PCF and SCOR share an axiom neither ever states: the performer of a process is internally assigned and owes its duty to the assigner. The firm's most valuable positions break exactly that — the performer is appointed by a court, a regulator or both disputing parties, and owes its duty to someone other than the payer.那些缺口里最硬的一簇——受监管行政、公允性意见、专家证词——缺的不只是节点。PCF 与 SCOR 共享一条从未写明的公理：流程的执行者由内部指派、并对指派者负有义务。而这家公司最值钱的一批位置恰恰全部打破它：执行者由法院、监管机构或争议双方任命，且对付款方以外的人负有义务。

A legislature once ran the control experiment: the UK Enterprise Act 2002 abolished “administrative receivership” because a receiver “has a primary duty of care to the appointing secured creditor, and that, in turn, may not result in the best outcome for the general body of creditors.” Same licence, same practitioners, same activities; one variable different — whom the duty runs to — and the legislature reached for the position rather than for higher professional standards.有一次立法层面的对照实验：英国 2002 年《企业法》废除了“行政接管人”，理由是接管人“对任命它的有担保债权人负有首要注意义务，而这未必对全体债权人产生最优结果”。同样的牌照、同样的从业者、同样的活动，只有“义务指向谁”这一个变量不同——立法者选择改造位置，而不是要求更高的操守。

The engine confirmed the layer is absent: six probes returned nothing above 0.634 (clean-anchor bar ≈ 0.70), and the sharpest one inverts. Asked how to demonstrate an assessor's independence to a regulator with verifiable credentials, the nearest concept in the framework is 11.2.2.9 “Maintain relationships with regulators as appropriate.” Where the position requires demonstrable distance, the framework offers relationship-building. The only top-ranked node whose name contains “independent” is 13.3.4.6.5 — where it denotes system access permissions. The word exists in the framework; the concept does not.引擎自己也证明了这一层不在框架里：六组探针全部不高于 0.634(干净锚点线约 0.70)，其中最锋利的一条是反向命中。问它“如何以可核验的凭据向监管机构证明评估方的独立性”，框架里最接近的概念是 11.2.2.9“视情况维护与监管机构的关系”。该位置要求可证明的距离，框架给出的却是关系经营。而唯一进入头部命中、名字里带“独立”的节点是 13.3.4.6.5——那里的“独立”指的是系统访问权限。这个词在框架里存在，这个概念不存在。

So the extension proposes a relation layer, not more nodes: who appoints, to whom the duty is owed, who pays, who can revoke, who holds a veto, and whether the qualification evidence is externally verifiable. The design rule is the point — independence is derived from these edges and cannot be declared by the modeller; modelling the claim explicitly is what makes it falsifiable. It ships as five executable SHACL constraints with a negative control: four samples that claim independence while the structure does not support it each trip the matching rule. The rules reject, they do not merely pass — which is the difference between a constraint and a slogan.于是扩展提案补的是一层关系语义，而不是更多节点：谁任命、对谁负责、谁付费、谁能撤销、谁有否决权，以及资格凭据是否外部可核验。设计准则才是要点——独立性由这些边推导，建模者不能自己声明；把“主张”显式建模，正是使主张可证伪的手段。这一层落成 5 条可执行 SHACL 约束，并配反向对照夹具：四个“声称独立、但结构不成立”的样本各自触发对应约束。规则会拒绝，不只是放行——这是它与一句口号的区别。

The transfer is the useful part: every operating company contains such positions — quality release, internal audit, a food-safety veto, a security sign-off. The pathology is identical: duty-owed-to ≠ reports-to, while the qualification is not externally checkable. The diagnostic runs directly on a client graph — enumerate every node holding a veto or release authority and test the slots for self-consistency. Where they fail, the finding is “this seat cannot work”, not “this person is not diligent”.可迁移之处才是有用的部分：任何企业内部都有这种位置——质量放行、内审、食品安全否决权、安全上线签字。病理完全相同：对谁负责 ≠ 向谁汇报，而资格又不可外部核验。诊断规则可直接跑在客户图上——枚举所有持否决权或放行权的节点，检查槽位是否自洽。不自洽的，结论是“这个位置无法成立”，而不是“这个人不尽责”。

Contributing back, and the honest boundary反哺产品，与诚实边界

The case also grew the role knowledge base outward from its IT origins — 24 professional-services roles (valuation, restructuring, forensic investigation, compliance, cyber risk, settlement / trustee / issuer administration…), all mapped to existing nodes. The KB moves to 86 roles and is live on the public query endpoint — a third party can verify it directly with list_roles / get_role_definition.本案还把角色知识库往 IT 之外扩——新增 24 个专业服务岗(估值、重组、取证调查、合规、网络风险、和解/受托/发行行政…)，全部映射到已存在的节点。角色库随之增至 86 个岗位，并已在公开查询端点上线——第三方可直接调 list_roles / get_role_definition 复核。

The engine gives the process scaffold, the audit trail and the coverage computation; this is an external, descriptive reconstruction, not a design commissioned by the firm. Weak or gap coverage is not failure but finding — it precisely marks the framework's extension boundary.引擎给的是流程骨架、溯源审计与覆盖计算；这是一次外部描述性重构，不是这家公司委托的设计。弱/缺覆盖不是失败而是发现——它精确地标出了框架的外延边界。

What compounds沉淀

A reusable method distilled into an asset: per-line engine chain · dual provenance · hollow-run check — turning the human-judgment question “does the framework fit a services firm?” into a computable coverage gradient plus a concrete extension backlog. Portable to any professional-services firm's process reconstruction.一条可复用方法论沉淀为资产：逐线引擎链 · 双重溯源 · 空转检验——把“框架适不适配服务业”这一人天判断，变成可计算的覆盖梯度 + 一份具体的扩展待办。可迁移到任何专业服务公司的流程重构。