A recent Enterprise Architecture Trends 2026 report from Avolution, a provider of enterprise architecture software, finds 92 percent of enterprise architecture leaders now name AI and agentic architecture as their top priority, with cybersecurity and risk management a close second, increasingly intertwined with AI adoption rather than treated as a separate concern. That figure, like most vendor commissioned research, is worth reading as a directional signal of where the profession is heading rather than a precise universal number. Even read that way, it describes something close to universal agreement that the core job of enterprise architecture has changed.
The same report projects that up to 50 percent of low level EA tasks, compliance checks, reporting and diagram generation among them, could be automated by AI agents by 2028. Together, those two findings describe a profession being reshaped from two directions at once.
What "AI Is the Top EA Priority" Actually Means in Practice
This is not simply enterprise architects adopting AI tools to do their own jobs faster, though that is happening too. It is enterprise architecture being asked to design for AI as a first class citizen of the enterprise, mapping how agentic systems consume and produce data, how they integrate with existing applications, and how their behaviour is governed at the architectural level rather than bolted on afterwards.
That is a materially different brief to the one most EA functions were resourced for two years ago, when architecture work was still primarily about applications, integrations and data flows that did not make autonomous decisions.
Why Cybersecurity and Risk Are Rising Together With AI
Agentic systems create new categories of architectural exposure that a traditional application inventory was never designed to capture: which agents have access to which data, what actions they can take autonomously, and how a compromised or misconfigured agent's blast radius is contained. Cybersecurity and risk ranking second, and rising alongside AI rather than independently of it, reflects architects recognising that these are now the same problem viewed from two angles, not two separate priorities competing for attention.
The Automation of Low Level EA Work, and What It Frees Architects to Do Instead
Automating up to half of an EA function's routine documentation and mapping work is, on its own, a genuinely useful efficiency gain. The more interesting question is what that frees architects to spend their time on instead. The honest answer, based on where the other 92 percent of priority is going, is governance judgement: deciding what an agentic system should be allowed to do, not just documenting what it currently does.
That is a higher value use of an architect's time, but it also raises the bar for what counts as valuable architecture work. Documentation and mapping were relatively easy to demonstrate the value of. Governance judgement on agentic systems is harder to make visible, even though it is considerably more consequential.
Why the EA Profession Itself Is Changing Shape
The same research notes that people from broader backgrounds are entering the enterprise architecture profession as the work becomes more multidisciplinary, and as AI makes architectural knowledge faster to surface and apply. A traditional EA career path built around years of systems documentation experience is no longer the only route into the profession, because the parts of the job that path was best training for are exactly the parts now being automated.
What This Means for How Organisations Should Resource Their EA Function Now
The practical implication is that an EA function still resourced and measured primarily on documentation output, system inventories, dependency diagrams, integration maps, is being resourced for the 50 percent of the job that AI is already automating. The organisations getting ahead of this are resourcing EA for the harder half: governing how agentic systems are designed, deployed and controlled across the enterprise, which is judgement work that does not automate away in the same manner.
What This Means for Your Organisation
What we see across the organisations we work with is that the EA functions adapting well to this shift are not the ones resisting AI automating their routine work. They are the ones that used the time freed up by that automation to build genuine governance capability over agentic systems, rather than simply doing the same documentation work faster.
Key Takeaways
- A fresh survey finds 92 percent of enterprise architecture leaders now name AI and agentic architecture as their top priority, with cybersecurity and risk management ranking a closely intertwined second.
- AI is already automating up to 50 percent of low level EA tasks such as documentation and baseline mapping, freeing architects for higher order governance work.
- Agentic systems create new categories of architectural exposure around data access, autonomous action and blast radius containment that traditional EA inventories were not built to capture.
- The EA profession is becoming more multidisciplinary as broader backgrounds enter it and AI reduces the value of experience built primarily around manual documentation.
- Organisations resourcing EA primarily for documentation output are resourcing it for the half of the job AI is already automating, rather than the governance judgement half that is not.
How Trusenta Can Help
Enterprise Architecture maps business capabilities, applications and integrations so agentic systems are governed at the architectural level this post describes, not documented after the fact.
Fractional Enterprise Architect gives organisations the governance judgement capability this shift requires without needing a full-time hire immediately.
Enterprise Architecture in the AI Era helps EA functions transition from a documentation focused operating model to one built around governing agentic systems.
Conclusion
Ninety two percent agreement on anything in enterprise architecture is rare enough to take seriously. What it describes is a profession being asked to govern agentic AI at the architectural level, not simply document it, at exactly the moment half its traditional workload is being automated. Organisations that resource their EA function for the harder half of that brief will be the ones actually ready for what comes next.
