- ArchEyes — 60,000+ architecture readers — published a full editorial on AI code compliance and featured Vitru as an example of where the category is heading.
- Avoidable design errors cost up to 21% of project turnover. That’s not a technology problem. That’s a workflow problem.
- There’s a critical difference between AI that reads code and AI that evaluates your actual model — most tools only do the first.
- Regulators are already using automated logic to check submitted drawings. The firms that pre-clear their models before submission will have a structural advantage.
- Vitru is built for model evaluation — checking real Revit elements against structured rules, with every finding traceable to a specific element ID.
We don’t write about press coverage often. But when ArchEyes — one of the most widely read architecture publications online — runs a full editorial on the problem you’ve been building to solve, it’s worth pausing on what they actually said.
The piece is a thorough look at AI code compliance for architects in 2026: where it works, where it doesn’t, what separates useful tools from noise, and what regulators are already doing. Vitru, our AEC venture, is featured as an example of the model-aware direction the space needs to go.
Here’s what matters from it.
The problem is bigger than most architects admit
Design errors are expensive. Not in the abstract — in actual project turnover.
These aren’t edge-case numbers. They represent the baseline cost of doing business in AEC the way it’s been done for decades: manual plan review, late-stage compliance checks, rework that shows up on site instead of in the model.
The window to catch a code error is early. The further it travels — from model to drawings to submission to site — the more it costs to fix. That’s not an insight. That’s just arithmetic.
Most “AI compliance” tools are solving the wrong half of the problem
This is the distinction the article makes that we think is genuinely important, and one that gets glossed over in most coverage of the space.
There are two completely different things that get called “AI code compliance.” The first is text interpretation — AI that helps you read and search the code. You ask a question, it answers, ideally with a citation. Useful, but limited.
The second is model evaluation — AI that checks your actual building model. The doors you drew, the egress paths you designed, the room sizes you specified. Checked against the rules. Flagged by element.
“An AI that reads code well is not the same as an AI that can check your model.”
Most tools are in the first category. They’re research assistants. Vitru is built for the second: querying Revit model data, running deterministic checks where rules are clear, and returning findings traced back to specific element IDs so the architect can act on them directly — not interpret a chat response and figure out what to do next.
The regulatory shift changes the equation
Here’s the part of the article that stuck with us most, because it reframes the urgency.
Regulatory bodies are already moving to automated plan review. Singapore’s CORENET X — mandatory for large projects since October 2025 — reportedly cut approval times by more than half by checking submitted BIM models automatically. Honolulu reduced reviewer time per plan from 60–90 minutes to 15–20. Austin, Los Angeles, and Seattle have live or committed deployments.
What this means in practice: the authority reviewing your submission is increasingly running the same kind of automated logic your tools should be running. If you’re not pre-clearing your model before submission, you’re essentially waiting for a machine to find problems you could have caught yourself — weeks or months earlier, when they were cheap to fix.
The industry term for this is “shifting left.” It means moving quality checks from the end of the process back to the act of designing. That’s been the core thesis behind Vitru from the beginning.
What Vitru does — and what it doesn’t claim to do
One thing the article is honest about, and we think it’s important to repeat: automation handles the prescriptive, quantitative half of a code well. Dimensional checks, clearances, egress widths, required properties, occupancy loads. That’s automatable today with high reliability.
The other half — performance-based provisions, judgment calls, anything requiring professional interpretation — is not. Won’t be anytime soon. The architects we work with know their code. Vitru is there to handle the checks that shouldn’t require their judgment at all, so they can spend it where it actually matters.
Vitru runs inside Revit. It reads model elements, runs checks against structured compliance rules and firm standards, and returns findings with the element ID, the rule, and the suggested fix. Every result is traceable. Every check is reproducible. It’s built around the professional’s judgment, not around replacing it.
We’re early. The agents are in beta. But the architecture firms we’re working with are already seeing real reductions in QA/QC issues and rework cycles — and the coverage in ArchEyes is a signal that the conversation around this is moving in the right direction.
Why this matters for ADAIA
Vitru is one of our ventures — built out of operator conviction that AEC is one of the sectors most underexposed to real AI infrastructure, and most in need of it.
The same principle behind everything we build at ADAIA applies here: the firms that automate the repeatable work earliest compound the advantage over time. The checklist your best QA manager runs in their head right now can become a rule every engineer runs on every model before it ever reaches review. That’s not a marginal improvement. That’s a structural shift in how a firm delivers quality.
That’s what we’re building toward. The ArchEyes coverage is a good marker of where the industry conversation is. The actual work is still ahead.
If you want to see Vitru on a live model, vitruai.com.