AI Can Generate the Diagram. It Cannot Generate the Consensus.

Eric Jager

Enterprise Architecture Author & Advisor

Published August 21, 2026 · 3 min read

For two years, AI enthusiasts have sold Enterprise Architects the promise that AI will take the modelling off their plates - the capability maps, the diagrams, the documentation - and free them to do the strategic work.

It was a comforting story. And it was only half true.

Enterprise Architecture faced several challenges well before generative AI arrived, one of which was that the products of EA teams were seldom utilized by the rest of the organization. It is crucial to understand that low usage does not necessarily mean low value - a single business scenario model that prevents a poor acquisition validates its worth, even if it is never consulted again.

That said, the promise of AI assumed that once production capacity is freed, strategic influence would naturally take its place. However, there is minimal evidence to support this belief. The hard part of this profession has never been drawing the boxes. It has always been the judgement underneath them - the trade-offs, the dependencies seen early, things that make a recommendation stick. AI can generate the diagram. It cannot generate the consensus.

So the real divide is not architects versus AI. It is architecture as an instrument versus architecture as a product. For the architects who always treated the models as a means to a better decision, this era changes very little; their value was never in the artifact. For the functions whose product was the documentation itself, AI will not feel like liberation. It will feel like an audit.

That is not an argument to slow down on AI. It is an argument to be honest about what your architecture practice is actually for. It is all about doing things right, instead of thinking you're doing the right things.