Is Your EA Team Ready for AI-Augmented Architecture? A 5-Factor Assessment

AI-Augmented Architecture readiness comes down to two things: the condition of the model and the skills of the team. We score them across five factors: a governed repository, a model with consistent structure that is complete and current, architects who can state intent, architects who can judge the result, and stakeholders who will use what the finished model produces. A license for a tool is not on the list. Access to a tool takes hours to set up. The five factors take months.

Most teams that start with the tool get one of two outcomes: nothing useful, or something that damages trust. AI that works on a poorly structured EA repository produces diagrams, mappings and answers that look authoritative and rest on unreliable data.

What AI-Augmented Architecture asks of a team

AI-Augmented Architecture is the way architects should be doing the job on Sparx EA: the architect decides what should be done, and AI does the mechanics in the tool and the model. Architects spend less time drawing diagrams, filling in properties and compiling reports, and more time deciding what the model should contain and whether what comes back is right. Once the model is finished and current, stakeholders get answers and views drawn from it without waiting for an architect to compile them.

That asks two things of a team. The model has to be trustworthy, structured and consistently governed, so AI has something sound to build on. The architects have to be able to state intent clearly and judge the result, because AI is only as good as the person directing it. For the bigger picture, see What Is AI-Augmented Architecture?

AI that builds on a poorly structured model produces more of the same problem, faster. Fix the structure and teach the skill, then let AI do the volume.

The 5-factor assessment

Score each factor from 0 to 4, then add them up. Factors 1 and 2 describe the model, factors 3 and 4 describe the architects, and factor 5 describes the people who will use the result.

• 0: Not in place
• 1: Partially in place, significant gaps
• 2: Mostly in place, some gaps
• 3: In place, minor gaps
• 4: Fully in place and maintained

Factor 1: Repository Governance

What it means: The Sparx EA repository has active MDG profiles, enforced element-type restrictions, mandatory tagged values, naming conventions, and a functioning review process. Packages are organized by domain and ownership is assigned.

How to assess it:

• Can you run a query against the repository and get complete, consistent results for Application Components with lifecycle status and owner?
• Does every top-level package have an assigned owner?
• Is there an active architecture review process for changes to baseline content?
• Have you scored 14+ on the Repository Governance Checklist?

Score 0-1: The repository has no structured governance, and the first job is to establish it. Score 2: Partial governance, and focused work will close the gaps. Score 3-4: Governance is sufficient to scale AI-assisted work.

Factor 2: Model Structure

What it means: The MDG profiles in use are correctly configured and consistently applied. Element types match the MDG standard types, tagged values are populated consistently, and relationship types are correct. The model is also complete and current for the part you want to work on, since AI can finish a model that is in shape and cannot guess at one that is not. (For why this matters so much, see MDG Technology as your AI quality gate.)

How to assess it:

• What percentage of Application Components have all three core tagged values, lifecycle status, owner, and criticality, populated?
• Are relationship types used consistently (Serving vs. Association vs. Composition)?
• Have you ever run an MDG validation report and resolved the findings?
• Is there a process to prevent new elements being created with incorrect types?
• When did the model last reflect what the project teams delivered?

Score 0-1: The MDG is installed but not governance-grade, and AI-built content would inherit the inconsistency. Score 2: Partial quality, and an assessment will identify specific gaps. Score 3-4: Quality is sufficient for AI to build on and for a person to check the result.

Factor 3: Stating Intent

What it means: Architects can say what they want in terms AI can carry out. Take navigation diagrams. The architect decides what experience a stakeholder should have when they open the model, where they should land and how they should drill down, and states that. Asking for “some diagrams” gets generic ones. The same skill applies to a mapping, a modeling standard or a model update.

How to assess it:

• Can team members write down the structure, level of detail and standard a model should follow, before anyone builds it?
• Do they know who will use each diagram or view and what that person needs to do with it?
• Has the team tried an AI tool on a real task in your repository, and kept what worked as a repeatable instruction?
• Has the team received formal ArchiMate training in the last 18 months?

Score 0-1: Without this skill, AI produces generic output that needs rework. Score 2: The skill is functional, and targeted practice will close the gaps. Score 3-4: The team can direct AI to a result it can use. (This is what the For Architects track builds.)

Factor 4: Judging the Result

What it means: Architects can look at what AI built and decide whether it is right. They recognize a wrong answer, trace it to the instruction or the data, and know where human judgment is required before anything is accepted. Automation checks completeness. People own correctness.

How to assess it:

• Can team members identify and correct model quality issues independently?
• Is there a review step before AI-built content becomes part of the baseline?
• Does the team share an understanding of where AI output needs a human decision before use?
• When AI finds an element with no stereotype, does someone decide what it should be?

Score 0-1: AI output would go into the model unreviewed. Score 2: A review habit exists but is uneven. Score 3-4: The team reviews and decides as a matter of course.

Factor 5: Stakeholder Pull

What it means: Business and IT stakeholders are familiar with the EA function, have received architecture outputs, and will use a finished, current model and the views built from it. Stakeholders who have never engaged with architecture outputs will not start using them because the model is better, however good it is.

How to assess it:

• Do you run a regular (quarterly or more frequent) architecture briefing for senior stakeholders?
• Have you delivered architecture-driven insights to a business stakeholder in the last 90 days?
• Is the EA function represented in project gate review processes?
• Have any stakeholders asked for direct access to architecture data?

Score 0-1: Relationships are not yet strong enough for a finished model to be used. Score 2: Relationships exist but are not mature, and a stakeholder engagement program is needed alongside the model work. Score 3-4: Stakeholders are ready to use what the model produces, through dashboards, reports or Copilot.

What is not on the list

Tool access is not a scored factor. A license and a few hours of setup get Claude, GitHub Copilot, Kernaro Assist or Kernaro AI Hub running, and AI Power Tools for Sparx EA is free from NovoCircle. Server-deployed options such as Kernaro AI Hub need a supported database backend, an HTTPS endpoint and, for Microsoft 365 Copilot, the right license level, and those are checked during setup. AI tools for Sparx EA compares which tool helps with which job.

Scoring guide and routing table

Total score = sum of the five factors (0-20). Use it to find your readiness state and the right next step.

Scoring guide and routing table

Score	Readiness state	Recommended next step
0-7	Not ready	Assess all five factors and agree where to start
8-11	Foundation work needed	Model structure and architect skills together, then reassess
12-15	Conditionally ready	Confirm gaps via assessment, then start one job and close gaps as you go
16-18	Ready to work this way	Set up the tools and apply them across the portfolio
19-20	Fully ready	Finish the model for the whole organization, keep it current, and put the finished model in front of stakeholders

The most common sequencing mistake

Teams frequently start with the tool. The reasoning is usually: “let’s see what the AI can do and then improve the model based on what it returns.”

AI tools do not reveal quality gaps in ways that build trust. They produce diagrams, mappings and answers that look authoritative. Stakeholders receive those and act on them. When the content turns out to rest on stale lifecycle statuses or incorrect owner tags, the damage lands on the credibility of both the architecture function and the AI work.

The sequence that works is to bring the model into shape and build the architects’ skills together, then let AI do the volume. An assessment that scores all five factors and produces a sequenced plan prevents the mistake. Paralysis to a Plan does exactly that, turning a five-factor score into a plan your management will fund.

Frequently asked questions

What does “AI-Augmented Architecture” mean?

AI-Augmented Architecture is the way architects should be doing the job on Sparx EA: the architect decides what should be done, and AI does the mechanics in the tool and the model. It asks two things of a team: a model in reasonable shape and architects who can direct the work.

What is the minimum model quality needed?

As a practical minimum, an MDG Technology active and consistently used for your repository (ArchiMate, for many teams), 80%+ of Application Components with lifecycle status populated, 80%+ with an owner tagged, and the primary capability-to-application relationships in place. Below those thresholds AI can still build, and the first job is closing the gaps. Anything that reads the model, including a dashboard or an assistant, reflects them.

How long does it take to become ready?

For teams scoring 8-11 (foundation work needed), model and skills work run together and usually take several months. Teams scoring 12-15 can start a first job within weeks once an assessment confirms the specific gaps, and close the rest as they go. The timing depends on how much of the model has to be built and how much practice the architects need.

What factor do most teams fail on?

Factor 2, Model Structure. Teams frequently have reasonable stakeholder relationships and capable architects, but have not enforced the model’s structure to the level AI needs to build on. Tagged values are defined but not mandatory, element types are mixed, and relationships are informal. This is the gap an assessment most often surfaces.

Should we train the architects before we start, or fix the model first?

Do both together. Architects who can direct AI but work in a model with no structure are limited in what they can do, and a well-structured model with architects who cannot state intent or judge results gets little use. Skills built on your own repository while the model is brought into shape is the effective sequence.

What does a readiness assessment cover?

A Sparx Services assessment scores all five factors: a repository governance audit (MDG configuration, access control, package structure), model structure and completeness scoring (element-type consistency, tagged-value completeness, relationship validity, how current the model is), an assessment of the architects’ skills at stating intent and judging results, and stakeholder relationship mapping. The output is a scored assessment and a sequenced roadmap.