How Enterprise Architects Can Use AI Today
Practical ways to improve thinking, content, and analysis—without new tools or setups
AI is currently discussed in enterprise architecture (EA) much the same way cloud once was: enormous potential, plenty of slides—and a sense that the real value is always just around the corner. The conversation often drifts toward premium EA tools with embedded AI features, custom language models, or carefully governed RAG environments that are not quite ready yet.
In practice, much of the value is already available. And it is available right now, using tools many architects already have open in another browser tab—ChatGPT, Copilot, or similar assistants. No new platforms. No integrations. No automations. You don’t need AI-enabled modeling tools or a carefully engineered environment to get started.
Used well, AI provides quiet but tangible benefits. It helps enterprise architects think more clearly, articulate ideas faster, and reach meaningful questions earlier. It lowers the threshold for getting started, improves the quality of architecture content and discussions, and offers a low-cost way to test assumptions before they harden into decisions. Over time, effort shifts away from mechanical work toward reasoning, communication, and judgment—the parts of EA work that actually matter.
This article is not about replacing EA tools, repositories, governance practices, or people. It is about using AI as an assistant in everyday architecture work. Below, I’ll walk through concrete, immediately usable ways enterprise architects can apply AI today—things you can try as-is, without waiting for the new AI strategy, the tooling roadmap, or the next budget cycle.
1. Sparring: AI as a Thinking Partner and Mirror
EA rarely fails because models are technically incorrect. More often, it fails earlier and more quietly. Focus gets locked in too soon. Assumptions remain implicit. Work drifts to the wrong level of detail, and conversations start only after decisions have already hardened. The result is architecture that is formally correct, but misaligned with what actually matters.
Used well, AI helps exactly here. Not by producing better diagrams or faster documentation, but by acting as a mirror for the architect’s thinking. AI does not do the architect’s work. It makes thinking visible before it turns into scope, roadmaps, deliverables, and decisions.
In practice, this kind of sparring is most useful at transition points: when you are about to start a piece of work, when scope feels unclear, when a deliverable keeps growing without improving decisions, or when you need to explain your thinking to others. A short pause and a few well-placed questions can change the direction of the work.
The easiest way to start is not to ask AI for answers, but to ask it how to begin. Take whatever you are working on—an EA deliverable, an assignment, a presentation, or a workshop—and ask simple framing questions: How should I approach this? Where should I start? What problem am I really trying to solve? That alone adds value, because it forces your starting point into words instead of leaving it implicit.
Over time, this turns into a habit. Instead of jumping straight into producing content, you begin by testing your approach and your explanations. Used this way, AI becomes a rehearsal space for architectural communication—revealing where reasoning breaks down and where meaning is lost, before you are in front of real stakeholders.
And something more fundamental happens as well. You are not just improving individual outputs. By repeatedly questioning how you start, frame problems, and explain your thinking, your overall way of working becomes more deliberate and reflective. In that sense, sparring with AI does not just improve the architecture you produce. It improves the architect producing it.
2. Content Creation and Refinement: Articulation, not Automation
Using AI for EA content creation is, at its core, very straightforward. You can ask AI to produce different kinds of EA content, and you can ask it to develop and refine content you already have. The output does not need to be technical or diagram-ready. In the time it would take to explain your modeling tool to AI, you can already model the result yourself in whatever tool you use. AI supports the thinking and articulation; you handle the modeling.
There are two equally valid starting points. You can ask AI to create content from scratch, or you can ask it to work based on existing material. In practice, the latter is often more powerful. Existing architecture documentation, workshop notes, interview summaries, and public examples usually contain most of the substance already. AI helps turn that raw material into clearer, more structured content that others can actually use.
It also helps to recognize that different kinds of EA content behave differently. Some content is definitional by nature. For example, capability structures and high-level data groupings are designed rather than discovered. There is room for interpretation, alternative structures, and naming choices. Here, AI works well as a partner for exploration: suggesting ways to structure things, spotting inconsistencies, and helping clean up abstractions. You remain responsible for what fits your organization, but getting started and exploring options becomes faster.
Other content is factual. Applications, integrations, data flows, platforms, and technologies describe what actually exists. This content must come from reality: documentation, repositories, and conversations with people. AI should not invent it. Where it adds value is in articulation—turning inventories into readable descriptions, explaining complex diagrams in plain language, and making models more consistent across the landscape.
The same logic applies to textual descriptions. Operating models, principles and their rationales, executive summaries, and workshop materials are often created under time pressure and refined later, if at all. AI helps produce an initial version quickly and then improve it iteratively, whether the starting point is a blank page or a pile of notes.
Over time, this changes how EA content is produced. Less effort goes into starting from scratch or wrestling with wording. More effort goes into reviewing meaning, improving clarity, deciding what actually matters, and refining and validation with stakeholders. Content creation and refinement become a continuous process—and architecture becomes easier to explain, discuss, and use.
3. Content Analysis: Sensemaking and Early Review
This use case sits between thinking and documentation, and it is often where EA work really starts: understanding what existing material actually tells you.
Enterprise architects handle collections of application lists, capability descriptions, process diagrams, and conceptual models. The first challenge is not redesign or optimization, but orientation. What is here? How clear is it? What can be understood from this as it stands?
At this level, AI works as a neutral reader. You can give it existing EA material and ask it to summarize what it sees, point out unclear areas, or highlight inconsistencies in terminology or structure. This helps improve clarity and shared understanding, but it stays at the level of description.
And that limit matters. As long as EA content is mostly structure—names, boxes, arrows, and lists—you can only say so much. You can make things look nicer, but you cannot yet say what is good or bad, important or trivial, expensive or cheap.
To get to substance, you need attribute data. Only when architecture elements are enriched with information such as cost, lifecycle status, quality, risk, or business importance does analysis become meaningful. That is when the question shifts from “what does this look like?” to “what does this mean?”
With attribute data in place, AI can help interpret patterns. It can highlight where cost and value don’t align, where critical areas rely on fragile solutions, or where cost and importance collide. These are not decisions or truths, but signals that help focus attention.
The distinction is simple but important. Clearer descriptions make EA material understandable. Better attribute data makes it decision-relevant. AI can support both, but only the latter supports prioritization and trade-offs.
AI still does not judge data quality, understand organizational history, or account for politics. It works with what it is given. Used correctly, it helps enterprise architects move faster from raw material to meaningful questions—without pretending to replace architectural judgment.
On Confidentiality, Abstraction, and Boundaries
Across all these use cases, the key question is not whether to use AI, but what level of input to use.
Many EA tasks work perfectly well with abstraction. Sparring, framing problems, drafting principles, structuring capability maps, or improving explanations can all be done without sharing confidential details. You can describe situations, approaches, and structures in generic terms and still get real value.
Analysis is different. If you want AI to help spot patterns, tensions, or improvement opportunities, it needs some real input—typically simplified architecture data such as an application list with a few basic attributes. Even then, the level of detail should be deliberate and minimal.
The practical rule is simple:
Do not share personal data, confidential business information, or anything you wouldn’t explain to a trusted external colleague.
Abstract first. Be specific only where it actually adds value.
When sensitive data is truly needed, use an environment that is appropriate for that data.
As always, the responsibility stays with the architect. AI does not decide what is safe to share or where boundaries lie—you do. Used with judgment, AI fits naturally into EA work without creating new risks. Used carelessly, it creates the same problems as any other poorly governed tool.
The Bottom Line
Using AI in EA is not about tools, it’s about starting. Start using what you already have, refine your way of working through real use, and gradually discover new use cases that fit your context.
This is not about cutting corners or outsourcing thinking. Quite the opposite. Used well, AI raises the bar. It helps you iterate faster, articulate more clearly, and examine ideas earlier and more critically. The goal is not to do less work, but to spend less effort on mechanical tasks and more on judgment, reasoning, and communication.
You don’t need a big plan or a perfect setup. Just begin, learn by doing, and adjust.
Even small steps—faster iteration, clearer writing, earlier analysis—add up quickly. And over time, they show up where EA actually matters: in better conversations, better decisions, and more effective architecture work.
🚀 Early Access: An AI Assistant Built Specifically for Enterprise Architects
I’m currently building an EA AI assistant as a custom GPT, designed specifically for enterprise architecture work: thinking, content creation, analysis, and everyday EA problem-solving.
The assistant is intentionally shaped around the same way of thinking about EA that runs through my books and articles. The goal is not to automate architecture work, but to support clearer reasoning, better framing, and more usable outcomes, in the same spirit as the work you’ve read.
You can get early test access before a wider release. All feedback—good, bad, and uncomfortable—is very welcome. The assistant will evolve continuously, as tools like this inevitably do.
It works with the free version of ChatGPT, but you’ll get significantly more value out of it with the paid plans and more capable language models.
If you’re curious to see what a purpose-built EA assistant can already do today, this is a good moment to try it and help shape what it becomes next.
The early-access version of the EA AI assistant is available here:
https://chatgpt.com/g/g-69284a2eedf8819199371368afbb866b-eetu-niemi-s-enterprise-architecture-assistant
🎄A Short Holiday Break
That’s all for now. This Substack will be on a short holiday break and return in early January.
Until then, seasonal greetings—and thanks for reading.
🎁 An Enterprise Architecture Book as a Practical Holiday Gift
If you’re looking for a useful, non-gimmicky gift for an enterprise architect—or for yourself—my book Enterprise Architecture – Your Guide to Organizational Transformation is designed exactly for that.
The book focuses on how EA actually works in practice: framing the right problems, staying at the right level of abstraction, and connecting architecture to real decisions. No tool worship, no framework overload—just clear thinking about what EA is for and how it creates value.
It works well as a holiday read, and even better as something to return to when planning and prioritization start again in January.
🔗 You May Also Like
Looking to dive deeper? Here are more enterprise architecture insights you might find useful:
👨💻 About the Author
Eetu Niemi is an enterprise architect, consultant, and author.
Follow him elsewhere: Homepage | LinkedIn | Substack (consulting) | Medium (writing) | Homepage (FI)
Books: Enterprise Architecture | Technology Consultant Fast Track | Successful Technology Consulting | Kokonaisarkkitehtuuri (FI) | Pohjoisen tie (FI) | Little Cthulhu’s Breakfast Time
Web resources: Enterprise Architecture Info Package (FI)
📬 Want More Practical Enterprise Architecture Content?
Subscribe to Enterprise Architecture Transformation for real-world advice on architecture that supports change, strategy, and delivery.
Spot-on about starting with what's already there! The sparring approach is underrated, AI as a thinking mirror is way more valuablethan template generation. I've found similar patterns in solutoin architecture where articulating the problem upfront saves hours of rework downstream. The distinction between definitional and factual content is clever too, keeps expectations realistic when dealing with existing landscape data vs conceptual structures.
Really interesting read. I put together a script to integrate "Archi" and each of the LLM engines so that Architects can push models directly to the chat bot and retrieve responses directly as "documentation" or "analysis" on the model. At the end of the day I guess it all comes down to the context and prompt you're using to deliver "value" from the model.
https://smileham.co.uk/2025/07/02/ai-support-for-archi-models-using-gemini-and-jarchi/