AI for the enterprise architect of today

Looking back on Boldo's session at Club Urba-EA, working workshop held on 28 May 2026.

A working session at the heart of the French EA community

On 28 May, the Boldo team was invited to present to Club Urba-EA, the reference association for enterprise architects in France. Active for more than twenty years, the association brings together 75 member companies and over 150 members around a purpose we fully share: "Growing together to transform better."

The workshop was not framed as a commercial demonstration, but followed a format the Club holds dear: a collective working session, designed to exchange, share experiences and build on collective know-how. The topic of the day could not have been more timely: the concrete contribution of AI to enterprise architecture work.

The opportunity was too good to stay theoretical. So we put forward a simple, almost provocative conviction: AI for the architect is not a matter of foresight. It is a matter of today.

The Boldo mindset: a metamodel that fits the company's mental model

Before talking about AI, we need to talk about foundations. And at Boldo, the foundation comes down to one idea: the tool must adapt to the organisation, never the other way around.

Too many architecture platforms impose a rigid formalism. ArchiMate everywhere, BPMN for everything, a single framework supposedly fit for every organisation. Our conviction is the opposite: one size fits none. The architect's first concern should be communicating with business colleagues, not debating with fellow architects in a methodological ivory tower.

That is why Boldo is built on a flexible, customisable metamodel, able to fit the mental model unique to each company. In practice, the architect builds their own language: their own layers, objects, properties and relationships.

And this metamodel does not sit on just any foundation. Boldo's technological core is a graph-based metamodel editor, structured on Neo4j. This choice is far from incidental, it directly serves the AI ambition:

• Depth of relationship traversal to explore the architecture in depth.
• Speed of impact calculations to answer concrete questions quickly.
• Native fit with LLMs through GraphRAG, because a well-structured graph is the ideal playground for contextualised AI.

Where the graph brings relational richness, structuring brings data integrity, fine-grained access management, workflows and history tracking. The best of both worlds, in the service of analysis.

AI for EA: a copilot, not a replacement

The message we were keen to convey to the Club is clear: AI does not replace the architect, it augments them. We talk about a copilot for the architect, present across the entire value chain.

Boldo offers three complementary integration modes, designed to fit the constraints of each organisation:

• AI by Boldo, with an open-source model hosted on Scaleway, to remain fully sovereign.
• Client AI (Bring Your Own Key), which lets you plug in your own Claude, GPT or Mistral key.
• The MCP protocol, to query Boldo from an external LLM portal and cross-reference architecture data with the rest of the IS (CMDB, Jira, data catalogue).

The chatbot and first agents launch officially in June 2026, with the MCP protocol following in July. Two structural advantages come up constantly in our conversations with architects: data sovereignty and AI cost control.

Concrete use cases, demonstrated live

It would not be a Club Urba workshop without something concrete. So we walked through several AI use cases in Boldo, live.

Creating views and dashboards. From a sometimes deliberately vague prompt, the AI builds a nested data-governance map, or composes a complete CISO dashboard: a pie chart of applications by criticality, an obsolescence radar by supplier, a risk map by impact × probability.

Automatic enrichment of the repository. The AI fills in asset properties and relationships in bulk, and offers autocompletion on records. Enough to dramatically speed up keeping the map up to date, a universal pain point for architects.

Explaining the architecture in natural language. The AI traverses the graph and explains the architecture around the CRM, identifies hot spots, and produces an actionable prioritisation matrix.

Analysing advanced use cases. The flagship example: analysing technological obsolescence risk. The AI detects that only a few technologies in the repository have an end-of-life date filled in, flags the governance gap, identifies the technical debt on the application side, and directly proposes the views and deliverables to produce in order to close the blind spot.

The common thread across all these use cases: the quality of the metamodel determines the quality of the AI. Garbage in, garbage out. The better the business and technical context is documented, the more the AI becomes a reliable partner.

A field testimony that rounds out the picture

The session was enriched by the testimony of an independent consultant and Core Data Model owner in a large home-healthcare group. His feedback on using Notebook LM, a RAG-type tool, to manage a corpus of around 90 business documents illustrated a shared reality: the difficulty of tracking decisions and aligning teams across multiple programmes. Further proof that AI's value emerges from the crossover between documentary search and contextualised generation.

Sovereignty and independence, running through every exchange

Unsurprisingly, in a room full of architects from major groups, data sovereignty ran through the entire discussion. Boldo addresses it by design: 100% European capital of which 99% French, governance handled by the founders, data hosted in France with Scaleway, GDPR compliance, and interoperability with no proprietary lock-in, guaranteeing full reversibility of the modelled data.

It is also this positioning that earned us, in April 2026, the number one spot in the G2 ranking of enterprise architecture solutions best rated by their users, ahead of the market's incumbents, particularly on ease of use.

What's next?

AI for the enterprise architect is no longer a promise. The Boldo chatbot launches in June, the MCP in July, and the roadmap continues with autonomous agents and the architecture-projection copilot, "Jarvis."

Club Urba-EA's AI initiative carries on too, with the next workshop on 30 June and a call for testimonies, formal or informal, to share the first lessons from the field.

A big thank you to Club Urba-EA for the warm welcome and the quality of the exchanges. Because at heart, this is exactly our conviction: growing together to transform better.

Want to discuss your own use cases for AI in enterprise architecture? Write to us at contact@boldo.io or find us at boldo.io.