Agentic EA

I admit to being a skeptic about how AI can be helpful for Enterprise Architecture. There's a great deal of discussion of various aspects of EA, including things like creating initiatives, prioritization, and oversight where I don't think today's models would be all that useful. (tomorrow may be radically different).

However, I do see value in agentic AI for EA. Today, most EA programs use an Architecture Review Board process, as described in various sources (including TOGAF). One limitation of the ARB structure is that it is necessarily limited by it's capacity. If there are 200 development efforts going on at one time, typically fewer than 10% can actually be reviewed by the ARB for compliance to standards or alignment to information strategy. That leaves a gap of 90% of efforts with no official guidance.

I think AI agents, trained on Enterprise Architecture standards and roadmap, can be helpful here. Any project team member can run an idea past an agent, at any time of the day or night, and get a rapid answer. No need to look up the EA sharepoint site and dig around to find out that the Boomi integration platform is required for objects representing real estate entries for example.

Agents can prepare documentation for the ARB, can review models, and can provide simple insights into existing capabilities or expected master data sources, making that information more freely available.

There would be requirements, of course, and a junior architect would need to be allocated to feed and update the core data used to train and retrain the agent. But this will not only free up senior resources but it should make it far easier for the developers and PMs involved in regular efforts to access EA roadmaps, thus making it far easier for folks to align.

Of course, this is a bit of a random thought and I may be way behind the times. By the time I post this article, an EA bot has probably already been implemented somewhere. I do think it could be useful though.