Stage 7. Running EA as a Capability: phase summary

6 min read 5 sections 7 key points

Stage 7 turns enterprise architecture from advice into a sustained operating capability. Across six modules you built the parts of that capability and the governance that keeps them working: the seven components of an EA capability from G184, the Architecture Board and its decision rights from G21I, compliance and waivers and contracts from G21H, the G198 skills framework and G249 maturity assessment, proportionate tailoring with G20F and G210, and the London Grid Distribution governance repository that holds all of it together. The thread running through every module is that structure on paper proves nothing; only visible behaviour and decision impact prove the capability is real.

Capability is an operating model, not a job title

G184 sets out seven components that together make an architecture capability: sponsorship and leadership, governance structure, method and process, content framework and repository, roles and skills, delivery interfaces, and measurement and continuous improvement. Hiring a chief architect supplies only the roles component, which is why the company in the opening case changed nothing in a year. The five diagnostic tests (sponsorship, decision, repository, delivery interface, exception) separate a working capability from an architect with a title. Architecture can sit in a centralised, federated, or embedded model, and most large regulated enterprises like London use a federated model or a centralised one with deliberate delivery embedding.

The Architecture Board protects coherence, not completeness

G21I defines the Architecture Board as the body responsible for the overall coherence of the architecture, so its job is to protect against cross-enterprise divergence, not to review every local design. The agency that reviewed 247 items and decided 11 failed because its scope was unbounded and its decision rights were assumed rather than written down. The fix is an explicit three-level decision-rights structure: the board decides, the domain architect decides with the board informed, or the delivery team decides within guardrails. A board of 6 to 10 senior members with a fortnightly cadence and an agenda that protects decision time stays selective; London runs an 8-member board whose decision log feeds Ofgem submissions.

Compliance has categories and layers, and contracts make agreements explicit

G21H gives compliance five conformance categories (irrelevant, consistent, compliant, conformant, non-conformant) so assessment never collapses into a binary pass or fail, and its review process runs through scope, evidence, assessment, findings, and report to the board. Beyond the rules, rule compliance asks whether formal standards and controls are met while intent compliance asks whether the architecture's deeper purpose survives local compromises; the bank in the case had perfect rule compliance and fourteen material deviations. An architecture contract is a joint agreement on deliverables, quality, and fitness for purpose, and a waiver makes a non-conformance visible, time-bounded, and reviewed rather than quietly accumulated.

Skills and maturity must diagnose, not decorate

G198 defines seven competency categories (generic, business, enterprise architecture, programme management, IT general knowledge, technical IT, and legal environment) across four proficiency levels (background, awareness, knowledge, expert), giving a realistic way to map who can do what. G249 assesses organisational capability maturity, but a maturity score is only useful when it answers the three-question test: what behaviour is weak, what consequence that creates, and what smallest improvement would change the operating reality most. For London the critical needs are OT/IT boundary expertise across SCADA and network automation, regulatory fluency in Ofgem and ED3, stakeholder facilitation, and repository stewardship, each carrying a concentration risk if held in one person.

Proportionate governance keeps TOGAF useful

Bureaucracy comes from poor implementation, not from TOGAF itself; C220 Part 3 says the ADM should be used flexibly and tailored. G20F provides four endorsed tailoring patterns (iteration and levels, phase selection, artefact selection, and governance scaling) and G210 shows how the ADM fits agile delivery through the architecture runway, sprint-aligned reviews, and decision-point governance without diluting the method. London applies three governance tiers (full weight, lighter, minimal) so the right governance cost lands on the right decision. The London governance repository ties everything together in seven sections (principles, target states, decision log, exception register, compliance records, architecture contracts, board charter), and the test is whether a delivery lead can actually use it to understand a decision.

Watch out for

  • Treating a chief architect appointment as an established capability when the other six G184 components are missing.
  • Letting the Architecture Board review everything, which clogs the agenda and means genuinely cross-enterprise decisions get decided casually or not at all.
  • Checking only rule compliance and assuming the architecture is protected, when a design can tick every box and still defeat its intent.
  • Reporting a maturity score as a finished output without explaining what behaviour is weak and what one improvement would change most.

Key takeaways

  • G184 names seven capability components: sponsorship, governance, method, content and repository, roles and skills, delivery interfaces, and measurement.
  • G21I makes the Architecture Board a coherence protector, with decision rights set at three levels: board, domain architect informed, or delivery team within guardrails.
  • G21H defines five conformance categories (irrelevant, consistent, compliant, conformant, non-conformant) and a five-step review ending in a board report.
  • Rule compliance checks formal standards; intent compliance checks whether the architecture's deeper purpose survives, and both are needed.
  • G198 has seven competency categories and four proficiency levels; G249 maturity is only useful when it diagnoses a weak behaviour, its consequence, and the smallest fix.
  • G20F gives four tailoring patterns and G210 integrates the ADM with agile sprints, letting London run three proportionate governance tiers.
  • London's governance repository has seven sections and proves the capability only through behaviour, not through a neat diagram.

With the capability, board, compliance, skills, proportionate governance, and the London repository now consolidated, the scenario practice puts them to work on realistic situations and the common mistakes before the timed stage assessment.

Start the scenario practice