1. Is Your EA Repository Ready?
An enterprise architecture repository should help answer a practical question. What is this agent allowed to do, for which business outcome, under whose authority, and with what evidence that it works? In this presentation, we explore the changes needed to support a multi-agent enterprise. We will connect business architecture to agent services, trusted data, approval controls, and operational evidence. The procurement example is a conceptual design, not a reported client implementation.

2. The repository connects design to execution
Application inventories and infrastructure diagrams still matter. But agents introduce additional dependencies: the tools they can invoke, the information they rely on, and the actions they can take across business domains. McKinsey describes an incremental modernization path that retains existing enterprise systems while introducing agent coordination. Microsoft highlights least privilege, task contracts, auditability, and human approval for high-impact actions. My recommendation is to make these concerns traceable in the repository, so architecture decisions remain connected to what is deployed.

3. Begin with business architecture
For supplier selection, the desired outcome might be faster evaluation while preserving purchasing controls. Supplier sourcing and selection is the capability. Gathering evidence, assessing risk, comparing options, and approving a commitment are the work. Business architecture identifies the accountable owners and the information needed for those decisions. Enterprise architecture connects that design to applications, data, services, and infrastructure. A capability does not automatically become one agent. Several agents may support one capability, or a simpler deterministic workflow may be sufficient.

4. Recommendation and commitment need different authority
Reading a supplier record, recommending a supplier, and committing company funds are different actions. The repository should distinguish them. A software agent can be represented as an application component, while the accountable human or organization remains a business actor or role. Document the permitted actions, approval conditions, escalation route, and operational fallback. An agent performing a task does not inherit accountability for the business decision. The architecture model records the boundary; runtime authorization and workflow controls must enforce it.

5. Procurement: evidence to approved commitment
In our procurement example, a supplier evidence agent retrieves approved supplier records. A risk agent evaluates the evidence against policy. An orchestrator combines the results into a supplier recommendation. The procurement approver reviews the recommendation and authorizes the commitment. Only then should the enterprise system create the approved purchase order. Each step has a different purpose and permission boundary. The downloadable ArchiMate model records components, functions, services, data objects, and approval responsibilities. Its preview simplifies routing; the relationship guide contains the complete register.

6. ArchiMate concepts for an AI profile
Use existing modelling concepts consistently. Goals, outcomes, capabilities, and value streams capture business purpose. Business actors, roles, and processes capture responsibility and work. Application components, functions, and services describe the agent-enabled application design. Data objects and artifacts distinguish logical information from deployed files and configuration. Requirements, constraints, assessments, and gaps support controls and change. These are proposed conventions using familiar ArchiMate three point two concepts, not an official AI extension. Check the version and relationship rules implemented in your repository before importing or changing definitions.

7. Trusted data and scoped tools
For each agent, identify the authoritative sources, their custodians, and the permitted read and write actions. Supplier evidence should come from approved records, with traceable provenance and freshness expectations. Tool contracts should describe allowed operations and required inputs. Keep secrets out of architecture records. MCP can support tool access, and A2A can support agent interaction, but neither settles business decision rights. The repository should connect integration contracts to the authority behind them. The control plane and operational systems must enforce those permissions.

8. Change impact and operational evidence
A model upgrade can change recommendation quality, cost, latency, or behaviour at an approval boundary. Before release, identify which workflows consume the affected agent, which policies and datasets it relies on, and which owners must approve the change. Link the deployed version to relevant evaluation results and operational dashboards. Keep full execution logs in their operational systems. The repository needs stable references for impact analysis, rather than a second copy of every log. Define who maintains each record and when it is reviewed.

9. Five questions for repository readiness
Can we trace every agent to a business outcome and a named owner? Can we identify its permitted read, write, and external actions? Can we locate authoritative data sources and their custodians? Can we see approval, escalation, and recovery responsibilities? Can we connect the deployed version to evaluation evidence and outcome measures? Use the answers to prioritize gaps. Then pilot one bounded workflow. Compare cycle time, quality, review effort, cost, exceptions, and control failures against a baseline before expanding autonomy.

10. One workflow before enterprise-wide expansion
Choose one business outcome and map its workflow, decisions, systems, and data. Extend the repository only with information that supports decisions and can be maintained. Validate the design with business owners and implementation teams. Test permissions, failure handling, and recovery before expanding the scope. The full article includes a practical readiness checklist, an illustrated procurement model, and an editable download. Visit Ahmad dot Alshami dot website for the article and presentation. Strategy connects to business capability, accountable decisions, agent-enabled execution, and measured outcomes.