Building Enterprise Architecture Artifacts with AI

Capability decomposition breaks a broad capability into more specific capabilities that describe work, not teams or applications. A capability earns further decomposition when its parts need separate treatment, such as being measured, funded, or improved on their own. If a decision would land the same way on the whole capability, decomposing further adds detail without adding value.

An as-is capability model records where a capability stands today, including its owner and maturity.

A to-be model records the target outcome the business needs, without naming a solution.

Comparing an as-is state to a to-be state is called a gap analysis.

A value stream shows how value flows through the business as stages of delivery and the capabilities that support them. A roadmap sequences change. A value stream follows one request, stage by stage, to the moment value reaches its recipient, often the customer.

Reading an application portfolio against the capability map reveals problems that span several systems or capabilities. From inside any one system, these problems are invisible. Common findings include:

• Data redundancy: systems keep separate copies of the same data with no reliable way to keep them aligned
• Gaps: a capability has work that no system covers
• Dependency chokepoints: several capabilities depend on one system, so its failure affects them all

An integration pattern is a standard way to connect systems. The choice starts from the shape of the interaction, meaning how many systems take part and how information flows between them. Factors such as timing, reliability, scale, security, and coupling also shape it. Each pattern trades one cost for another, so architects add infrastructure like a broker or gateway only where an interaction needs it.

An interface catalog records each interaction between applications, including the information exchanged and the integration pattern selected for it. With every entry in one place, anyone planning a change can see which systems it will affect.

Data governance defines the policies every system must follow.

Data ownership names who is accountable for a dataset. Systems can follow the same policies and still disagree, and the owner decides which record to use.

Data stewardship is the day-to-day work of keeping data usable, such as merging duplicates, handling access requests, and checking for drift between systems.

An ownership charter documents a dataset’s owner, steward, and escalation path. It establishes who is accountable for the data, not which system holds the trusted copy. The escalation path gives a disagreement somewhere to go when the owner and system teams cannot settle it. Many organizations record the same assignments in a responsibility assignment matrix.

A system of record is the authoritative source for a business entity or data element. Common criteria for choosing one include where the data is first captured, where it is corrected, which system holds the most current version, and which system has the accountability to govern it. One system can govern a whole entity while another governs a single element of it. When copies conflict, the other systems defer to the system of record.

A trade-off matrix compares platform options against shared decision criteria, such as cost, control, scalability, and compliance. Each option usually wins on some criteria and loses on others. Binding criteria, or must-haves, rule out any option that fails them. The choice among the options left depends on which trade-offs the organization can accept.

An Architecture Decision Record (ADR) typically captures the context that required a decision, the options considered, the decision made, and the rationale behind it. It also records the consequences, meaning what the decision commits the organization to, and a status that shows whether it still applies. Recording the options that were set aside, not just the one chosen, lets a future team evaluate the decision instead of re-deriving it.

A gap analysis compares a current state with a target state. Across the whole capability map, it finds capabilities the business needs but lacks. For a single capability, it compares the as-is state with the to-be state and sizes the change needed to close the gap.