AI can help list, group, trace and explain facts already recorded in a capability model: hierarchy, owners, maturity assessments, strategic links, priorities and related applications or programs. It cannot reliably answer questions about relationships or assessments the model does not contain.
The best questions name the model, scope, evidence fields and desired output. Ask the AI to cite the capability records used, separate missing data from negative findings, and avoid inventing causes or recommendations.
Disclosure: I’m Nick Malik. I built Useful EA, a commercial set of architecture tools, and the section on Useful EA below describes it. The method in this article works with any repository and any AI assistant.
Start by asking what the model actually contains
PromptDescribe the capability-model data you can access. List the available hierarchy levels, attributes, assessment fields and relationship types. State which fields are incomplete. Do not infer unavailable relationships.
This prevents an AI from answering a strategy, application or cost question using data that exists only in your expectations.
Eight useful capability questions
| Question to ask | Data required | Safe interpretation |
|---|---|---|
| Which L3 and L4 capabilities roll up to Customer Management? | Parent and child hierarchy | The recorded descendants, not an industry-standard decomposition |
| Which capabilities support the Growth strategy? | Capability-to-strategy links | Recorded support relationships, not proof of realized contribution |
| Which capabilities have the largest maturity gaps? | Current and target assessments plus calculation rules | A ranking under the recorded scores and rules |
| Which high-priority capabilities lack an owner or recent assessment? | Priority, owner, assessment date | Data-quality exceptions requiring follow-up |
| Which applications support Order Management? | Capability-to-application relationships | Recorded application support, not complete operational dependency |
| Which capability names may overlap or be ambiguous? | Names, descriptions, parents | Candidates for human review, not confirmed duplicates |
| Which high-priority gaps are not covered by a proposed program? | Priorities and gaps plus program scope | Recorded planning gaps, not proof that no work exists elsewhere |
| What should stakeholders verify before funding this improvement? | Scores, strategy links, owners, source notes | A review checklist, not an investment decision |
Prompt 1: explain a capability hierarchy
PromptFor the capability named Customer Management, list every recorded descendant in hierarchy order. Show capability number, name, parent and level. Flag missing parents, duplicate numbers and children whose level cannot be determined. Do not add capabilities from general knowledge.
Verify the result against the model itself or its exported hierarchy. If the model allows more levels, ask for all of them rather than assuming only L1 and L2.
Prompt 2: find maturity gaps without hiding missing data
PromptRank assessed capabilities by the recorded current-to-target maturity gap. Show the component scores and the calculation used. Put capabilities with missing current or target values in a separate "not assessable" list; do not treat blanks as zero.
"No recorded score" is not "no gap."
Prompt 3: connect capability gaps to strategy
PromptFor each capability linked to the Growth strategy, show the recorded relationship, current maturity, target maturity, gap and owner. Separate direct links from inferred relevance. List capabilities with no assessment separately.
An AI may summarize the relationship, but it should not claim that improving the capability will cause a strategic outcome unless the model and evidence establish that logic.
Prompt 4: identify planning omissions
PromptList high-priority capability gaps that are not assigned to any proposed program. For each, show the priority inputs, related strategy and accountable owner. Also list programs claiming the same gap so duplicate coverage can be reviewed.
The result is an investigation queue. Work may exist outside the model.
Prompt 5: prepare a stakeholder review
PromptCreate a review brief for the ten highest-priority capability gaps. For each, state the recorded evidence, missing data, owner, linked strategies, proposed program coverage and the questions stakeholders must resolve. Cite every model record used.
This is a good use of AI: turning structured facts into a readable agenda while keeping the uncertainty visible.
What the model cannot answer alone
Do not ask a capability model to prove:
- that investment will produce a business result;
- that every application or process supporting the capability has been recorded;
- that a high maturity gap is more urgent than an unmodeled risk;
- that similarly named capabilities are duplicates; or
- that a missing relationship means no real-world relationship exists.
Those conclusions need owners, operational evidence and decisions beyond the model. For how capability models support planning in the first place, see practical uses of a business capability model.
Where Useful EA fits
Claude, ChatGPT, Claude Code, Cursor and VS Code can connect to Useful EA directly at mcp.useful-ea.com. You sign in with your usual Useful EA account, and the assistant reads only what you can already open, with your permissions; it can’t change your models. A company’s administrator can turn assistants off for the company’s members. How to connect an AI assistant.
Connected to Capability Modeler, an assistant can list the models you can open and read one model’s capabilities with their maturity, target, cost and stewards, its strategies and its views. It can also read the model’s priorities report, which ranks capabilities by the gap between current and target maturity, strategy alignment and investment. Applications linked to capabilities live in Application Portfolio Manager, which the same connection can read.
If you’d rather work from a file, export the model as CSV. The export holds each capability’s number and parent, steward, current and desired maturity (overall and for people, process, tools and information), cost and investment fields, linked strategies, views and calculated priority. Proposed programs export separately from the Program Board.
Only give architecture data to an AI assistant your organization has approved for that information. An export can include names, owners, costs and plans that are confidential.
Practical rule
Ask AI to explain what the capability model says, expose what it does not say and prepare the next human review. Do not ask it to turn a recorded architecture relationship into proof of business value.
Sources
Vendor documentation is cited as context for how the market describes these patterns, not as evidence about Useful EA.
- Ardoq: Omnipresent AI Assistant — natural-language repository questions, document extraction and changes staged for authorized review
- Ardoq: Q1 2026 AI roundup — questions such as which applications support a capability, and diagrams converted into candidate data
- SAP LeanIX: MCP server for enterprise architecture — structured AI access to application records, capability maps, decisions and relationships