Here's what we hear consistently from EA leaders: their teams are brilliant, strategic people who are stuck doing work that shouldn't require them.
Manually populating application data from contracts. Writing capability maps from scratch. Cross-referencing spreadsheets to answer a question someone could have phrased in plain English. Responding to the same "do we already have a tool for this?" question from eight different teams.
That's not architecture. That's administration.
We estimate that 40% of data-collection EA tasks can now be handled by AI agents, and we're not talking about a hypothetical roadmap item. Tenneco is already there. Abbey Cletus, Head of Strategy and Automation at Tenneco, put it plainly:
"It feels like we're multiplying the team without multiplying headcount."
Their goal: agents handle 40% of the EA team's workload in 2026. They built six AI agents embedded in Microsoft Teams Copilot via Ardoq's MCP Server. Estimated ROI: 292%. That's a tracked production outcome. Not a demo.
The question isn't whether agentic AI belongs in EA. It's whether the agents reasoning on your data are actually reading your data — or someone else's best guess at it.
Every Ardoq AI agent reasons directly on your live architecture graph — the proprietary technology that maps every application, dependency, capability, and risk in real-time. Not a stale export from 3 months ago. Not a PDF someone uploaded. Your actual, current architecture.
That distinction matters more than it sounds. When an LLM must retrieve ten or more interconnected facts correctly in sequence, accuracy drops to 43%, roughly coin-flip odds. Ardoq offloads those calculations to deterministic code. Our AI Assistant calculations are computed, not guessed. When you ask "which 30 applications are in scope for my cloud migration?" — you need the right 30. Not 27 correct with 3 hallucinations.
And critically, no AI agent in Ardoq ever commits a change directly to your live architecture. Every output goes to a Scenario branch first. Human review is required. That's not a feature we added, it's the bar we built.
Previously, the AI Chat Assistant worked with our reports only. Useful, but narrow. The Omnipresent AI Assistant follows you through the entire platform — Dashboards, Viewpoints, Metamodel, Reports — and understands the context of where you are and what you're looking at.
Ask it to “map all CRM applications to our Customer Engagement capabilities” or to “Draft a value stream for Order-to-Cash using the applications and capabilities already in Ardoq” Ask it to “Read these SaaS agreements and flag any applications that aren't already in our Ardoq inventory. Mark the owner and renewal date for each”. It surfaces answers grounded in your architecture, not the internet. And because it's context-aware, questions asked while you're looking at a specific application automatically prioritize that asset before searching wider.
This is also the primary interface for triggering all of Ardoq's other AI agents. One front door to your entire agentic workforce.
Try our new omnipresent assistant.
Getting data into EA models has always meant months of manual work that no one has time for. Contracts arrive as PDFs. Application details live in spreadsheets. Vendor information is buried in email chains. Important sketches are done on whiteboards.
The Data Ingestion Agent changes the entry point entirely. Drop in a software contract and the agent reads it — extracting the vendor, lifecycle stage, renewal date, SLA terms, application name — and maps it straight into your Ardoq model. Describe something in plain English and the agent handles the data entry. Connect your own external discovery agents via MCP and they can write back into Ardoq automatically, keeping your architecture current without manual handoffs.
Governance stays in place throughout. Nothing lands in your live architecture without human approval.
Your architecture data has gaps you don't know about yet. Applications without owners. Capabilities without mappings. Risk signals buried in a dataset too large for manual review.
The Foundation Insights Agent analyzes your Foundation architecture data continuously — spotting unusual patterns, flagging data quality issues, and surfacing the things worth a conversation before they become incidents. It doesn't make decisions or change data. It tells you what to look at next, explains why it matters, and suggests who to talk to or what to check.
The goal is the same as any good EA practice: the right conversation at the right time, before the problem becomes a fire.
See the Foundation Insights Assistant in action.
Ardoq's out-of-the-box solutions — Application Rationalization, Technology Risk, Business Capability Management — are purpose-built for the work EA teams do most. The Agents in OOTB Solutions bring AI into those workflows directly, handling the time-intensive tasks so architects focus on the decisions those tasks were always in service of.
Three agents ship initially:
With many more scheduled to go live throughout 2026. The work your team shouldn't be doing manually — agents handle it.
Learn more about Ardoq's new AI features for generating value streams, capability maps, and governance checkpoints in seconds.
This one matters for the EA teams who've outgrown what off-the-shelf AI can do.
Custom Agents lets Enterprise Architects build and deploy their own AI agents inside Ardoq, scoped to their metamodel, their governance rules, and their specific workflows. No code required. Define what the agent knows, what tools it can use, and whether it has read-only or controlled write permissions. Share agents across the org like you share reports — with viewer or editor access.
This is the part that makes Ardoq meaningfully different from any other AI play in the EA market. Unlike generic AI layered on top of your data, Custom Agents are configured around your architecture. An agent that knows the difference between a critical dependency and an orphaned application. One that understands your specific metamodel, not a generic template.
Tenneco built six. You can build the ones that fit your landscape: learn more.
Beyond the agent layer, Ardoq is shipping a set of capabilities that change the underlying experience of working in the platform.
AI Import Builder connects any third-party data source to Ardoq in minutes. An AI custom integration builder, essentially. The AI reads third-party API documentation and configures the connection automatically — no scripting, no engineering dependency.
AI Semantic Search makes Ardoq a frontier EA platform with context-aware search that understands meaning, not just keywords. Ardoq will surface better search results from meaning, not only keywords. This means that business users can search in their own words and find what they need, even without knowing the exact asset name.
AI Query Builder takes advanced search from a specialist capability to something any stakeholder can use. Type what you're looking for in plain language, and Ardoq builds the query. No syntax. No expertise required.
Safe Data Population expands what the AI Assistant and MCP Server can do — not just reading your architecture, but safely writing back to it. Every population action follows your approval workflows.
MCP Single Sign-On brings OAuth authentication to Ardoq's MCP Server, replacing the manual API key management that every team using Ardoq MCP currently deals with. One sign-on. Full admin control. Automatic token refresh. The same authentication standard your org already uses for Salesforce and Workday.
AI Web Search Tooling pulls current, real-world information from the web and brings it into your architecture work — enriching your model with vendor data, application reviews, or market signals without manual research. Still under evaluation for security and governance implications, but a significant capability for keeping architecture models relevant to the world they're operating in.
The EA teams who get this right in 2026 are going to look very different from the ones still running on spreadsheets and quarterly surveys.
They'll have architecture models that update themselves. They'll have agents handling the data work so architects can show up to board meetings with traceable recommendations instead of apologies about data quality. They'll be the team that makes the CIO's AI investments actually work — because without a current, connected architecture graph, every Copilot in the business is reasoning on stale data.
The 40% figure isn't a marketing claim. It's what Tenneco is targeting this year, with agents already in production.
If your EA team is still doing work that AI can handle, that's not a capacity problem. It's an architecture problem. And that one we can fix.
Explore Ardoq's AI capabilities or book a demo now to see the leading AI-first EA platform in action.