Governed architecture knowledge
Capability, value-stream, stakeholder, information, organization, product, strategy, initiative, policy, metric, decision, and evidence records.
SCALE
Strategic Capabilities Architecture Linkage Engine.
An open reference design for an AI assistant that can retrieve, link, test, and explain governed business architecture knowledge without pretending to know what the enterprise has never documented.
The problem SCALE addresses
Business architecture repositories are difficult to use even when the content is good. The knowledge is distributed across models, spreadsheets, decks, decisions, SharePoint libraries, architecture tools, portfolio systems, and the memories of experienced practitioners.
Traditional search finds documents. A useful architecture assistant must understand typed objects and relationships: which capabilities enable a value stage, which initiatives change them, who owns the decision, what information is involved, how success is measured, and which source supports the answer.
SCALE is the design for that linkage engine. It is not a claim that an AI model can infer the enterprise from generic knowledge. The architecture must be governed before the assistant can be trusted.
Reference architecture
Six layers from source to governed answer.
The model separates knowledge quality, linkage, retrieval, explanation, and governance so a compelling interface cannot hide a weak repository.
Canonical vocabulary and identity
Resolve duplicates, definitions, levels, ownership, status, effective dates, provenance, and confidence before retrieval.
Typed architecture relationships
Connect objectives to outcomes, value streams, capabilities, initiatives, information, owners, measures, risks, and enabling technology.
Question-aware orchestration
Interpret the decision behind the question, select relevant domains, traverse relationships, and return the smallest useful evidence set.
Source-linked responses
Show the answer, reasoning summary, cited records, conflicts, gaps, confidence, and the owner who can resolve uncertainty.
Human-controlled change
Keep AI read-oriented by default. Proposed additions or changes enter normal stewardship, review, versioning, and decision processes.
Example questions
Questions a linkage engine should answer.
Each answer should include the architecture path it followed, its sources, what is missing, and where human judgment remains necessary.
What changes if this initiative moves forward?
Trace from initiative to value stages, capabilities, information, stakeholders, owners, measures, policies, risks, and enabling systems.
Which capabilities must improve to achieve this objective?
Show the linkage, current performance evidence, active investments, gaps, overlap, and unsupported assumptions.
Who can decide this, and who is merely affected?
Distinguish capability stewardship, organizational authority, stakeholder interest, architecture governance, and delivery accountability.
Have we solved or modeled this before?
Find prior scenarios, decisions, patterns, approved definitions, rejected options, and the conditions under which they remain valid.
Where is the architecture internally inconsistent?
Detect duplicate names, orphaned elements, broken relationships, stale sources, missing owners, and contradictory classifications.
Why is this a capability rather than a process?
Teach from the enterprise’s own approved examples while linking back to formal definitions and local modeling rules.
Design principles
Guardrails are part of the product.
The most dangerous failure is not an obvious error. It is an answer that sounds institutionally informed but silently blends stale facts, inference, and generic knowledge.
Evidence before fluency
A polished answer without traceable enterprise sources is a liability. Every material claim should point to a governed record or be labeled as inference.
Links are the product
The value is not another document search. SCALE should follow relationships across the architecture and explain why they matter to the question.
Decision-centered retrieval
The engine should ask what decision is being made and avoid flooding the user with every related artifact.
No silent invention
Unknowns, conflicting definitions, stale records, and absent ownership should be exposed rather than filled with plausible text.
Architecture remains accountable
Named stewards and formal decision rights remain responsible for enterprise knowledge. AI does not become the owner.
Start narrow
A useful pilot can cover one value stream, its enabling capabilities, active initiatives, owners, measures, and source artifacts before enterprise expansion.
Build in public
Use the design, challenge it, or contribute a use case.
SCALE will develop through reference schemas, prompt and retrieval patterns, governance rules, example questions, and small implementation experiments.
Have a repository, platform, or use case?
Send the sanitized problem—not proprietary architecture content.
Email a SCALE use case