Insight · AI-Augmented Architecture

What an Architect Can Hand to Kernaro Assist in Sparx EA, and What Stays Theirs

By Sparx Services · June 17, 2026 · 7 min read

Kernaro Assist is an AI assistant built into the Sparx EA client, from Sparx Systems. It takes the mechanical weight out of modeling: creating elements, drafting diagrams, checking the model against your standards and answering questions about what is in the repository, without leaving the tool. The architect keeps the decisions, the stakeholder conversations and the quality of the model. Released in May 2026, it is new, so treat what it produces as a strong first draft.

The AI works from what you tell it and what is already in the model. The quality of both sets the quality of the result.

Two things decide how well this goes. The first is the person giving the instructions: AI is only as good as the person prompting it. The second is the MDG Technology and the repository underneath. An architect working in a well-governed repository saves real time. An architect in an ungoverned one gets suggestions that are incomplete and inconsistent, because there is no coherent model to read.

What an architect can now get done with Assist

Each block below follows the same pattern: how the task is done today, how it is done AI-augmented, and the skill the architect needs.

Getting new elements into the model

Today. The architect clicks through the UI, creating each element, placing it in the right package and filling in tagged values one field at a time. Bulk creation is slow enough that it often does not happen.

AI-augmented. The architect describes what is needed and Assist creates it. “Create a Capability element called Customer Onboarding under the Customer Engagement domain package, with MaturityLevel tagged as 2 and InvestmentCategory tagged as Transform” produces the element in the right repository package with the tags populated.

The skill. Stating the intent precisely: the type, the package, the tags and the naming. A vague request produces a vague element. Precision comes from knowing your own modeling language well enough to say what you want.

Getting a first draft of a diagram

Today. The architect starts from a blank canvas and places every element and connector by hand.

AI-augmented. The architect gives Assist a scenario: “Create an ArchiMate diagram showing the Customer Onboarding capability, the three applications that realize it, and the data entities they consume.” Assist drafts a structure with suggested connectors, and the architect works from there instead of from nothing.

The skill. Knowing what the diagram is for and who will read it, then reviewing the draft for connector types, placement and anything misread. The draft is a starting point. Whether it represents the architecture accurately is the architect's call.

Checking the model against your standards

Today. Someone reads through large packages looking for missing tagged values, elements built on base metaclasses where a stereotype should apply, and names that break convention. It is slow, so it happens rarely.

AI-augmented. Assist checks a package or a set of elements against the MDG profile and flags what does not conform.

The skill. Writing the standard precisely enough to check, and deciding what to do with the exceptions. A check can say a field is empty. Whether the value in a field is right remains a human judgment.

Answering a question about the model

Today. The architect builds a matrix or writes a query to find, say, applications with a lifecycle status of Under Review and at least two capability realization relationships.

AI-augmented. The architect asks the question in plain language and Assist traverses the repository and returns structured results.

The skill. Asking a question the model can answer, and judging whether the answer is complete. If tagged values are sparsely populated, “which applications are at end of life?” returns only what happens to be filled in, which might be a fifth of the real answer, and misleading because it looks complete.

Filling in tagged values consistently

Today. Blank fields stay blank because filling them in is tedious, and the ones that get filled drift apart in wording.

AI-augmented. When new elements are created, Assist suggests tagged-value completions drawn from the MDG profile and from how similar elements are already filled in.

The skill. Deciding which values are required and what each one means, so the suggestions have something consistent to follow.

Why the MDG matters so much

Assist reads the repository to produce everything it returns. Its suggestions about element types come from the stereotypes defined in your MDG profile, and its answers come from the tagged values and connectors that are populated. Nothing it returns is better than what it reads.

If your MDG profile is undefined or applied inconsistently, Assist cannot infer what you mean by “application.” It sees a mixture of ArchiMate Application Components, UML Class elements pressed into service as application placeholders, and free-text notes. It cannot produce a coherent list of all applications because the repository has no coherent concept of one.

The same constraint applies to every AI that reads or writes the model, including Kernaro AI Hub, Claude or GitHub Copilot with AI Power Tools, and every report built on EA data. We make the same point at length in why your EA repository quality determines your AI output quality. If your governance is weak, the best first use of Assist is governance improvement itself: conformance checking, tagging backfill and element correction.

What stays with the architect

Architectural decisions. Whether to retire an application, whether a capability is strategically important, whether a given integration pattern fits. Assist can surface the information behind a decision, such as what depends on an application and what its maturity rating is. The decision stays with the architect.

Stakeholder engagement. Architecture is a communication discipline. Understanding what business stakeholders need, translating that into architecture and presenting findings people can act on is not automated. Assist cuts the mechanical overhead and leaves more time for that work.

Context the model does not hold. Which of three identically named applications is the canonical one, and what lifecycle status belongs on a system whose owner has left, are answered by asking people.

The limits Sparx states. The Basic license is read-only, with no agents and no central audit trail. Read and write access and agents come with the Advanced license, and shared agents and the audit trail need Kernaro AI Hub. Check your license against Sparx Systems' table before you plan a team rollout.

Which helper fits which modeling job

Assist is one way to get modeling work done with AI, and the right choice depends on how your architects work.

The architect is at the desk in the EA client and wants to work in a panel beside the model. Kernaro Assist.
The architect wants to state a larger piece of work in a chat and have the AI carry it out on the repository, such as navigation diagrams for every application or a mapping pass across the portfolio. Claude or GitHub Copilot with AI Power Tools for Sparx EA, which is free and published by NovoCircle.
Stakeholders need to ask questions across the repository and other tools. Kernaro AI Hub.

The AI tools for Sparx EA comparison puts all six options side by side.

What the time goes to

In most EA teams a meaningful slice of the day goes to creating elements by hand, populating tagged values one field at a time, building routine diagram drafts from a blank canvas, checking packages for conformance and compiling answers to ad hoc data questions. When AI takes those on, architects get time for reading what the business needs, making trade-offs, engaging senior stakeholders and designing patterns that hold up. We unpack where the hours go in what 70-80% of architect time actually goes on.

What the team does with the recovered time decides whether the tool delivers value. Time poured back into more mechanical modeling is a marginal gain. Time redirected into design and stakeholder work is a shift, and it is the shift behind AI-Augmented Architecture.

Putting Assist to work

Three things should be in place before you switch it on.

Set the MDG governance baseline

Assess and strengthen the MDG profile first: define stereotypes, establish tagged-value schemas and run a conformance pass over existing content. This sets the ceiling on what Assist can produce.

Teach people to prompt and to review

Decide which tasks go to Assist first, which stay human-led and which need review before publishing. Then practice the instructions: architects who state type, package, tags and purpose get better drafts than architects who describe a vague wish.

Build a feedback loop

Give architects a simple way to flag incorrect or unhelpful output. Errors that cluster often point to a gap in your MDG, and fixing the MDG improves every later draft.

Frequently asked questions

What is Kernaro Assist?

Kernaro Assist is an AI assistant embedded in the Sparx EA client. Architects work with it through a natural language panel to create elements, generate diagram drafts, run governance checks, and query repository data without leaving the EA environment. It is a Sparx Systems product, released in May 2026, and it is bring-your-own LLM: your organization supplies the keys and pays the provider for usage.

Does Kernaro Assist need Kernaro AI Hub?

No. Kernaro Assist runs inside the EA client and works standalone on the Basic or Advanced license. The Basic license is read-only, and the Advanced license reads and writes. Connecting the Advanced license to Kernaro AI Hub adds centralized access control, shared agents and audit. The Hub is the web layer for cross-tool questions and stakeholders; Assist serves the architect at the desk.

What is the MDG dependency for Kernaro Assist?

Kernaro Assist's output quality tracks the MDG governance quality of the repository it reads. A repository with defined stereotypes, consistent tagged values, and populated properties produces accurate, useful responses. A loosely governed repository produces incomplete and inconsistent ones. Strengthening the MDG profile before deploying Assist is the most reliable way to raise its value.

Can Kernaro Assist create full ArchiMate diagrams from a description?

Assist generates first drafts of ArchiMate diagrams from a natural language description. It creates elements, places them, and draws suggested connectors. The result is a starting point, not a finished diagram: architects review and refine connector types, validate placement, and confirm the diagram represents the architecture accurately.

Is Kernaro Assist a replacement for architectural judgment?

No. Assist handles mechanical modeling tasks: element creation, diagram drafting, governance checking, repository queries. It does not decide whether to retire an application, which integration pattern fits, or whether an investment is justified. Those calls require business context and judgment Assist does not hold. It supports architects and does not substitute for them.

We can teach your architects to prompt, review and govern AI-assisted modeling on your own repository, or build and update the model for you and hand it over.

Get the modeling done, and keep the judgment with your architects.

We can teach your architects to work this way on your own repository, or build the model for you and hand it over. Most teams want some of both.

Book a call

Keep reading