The architecture, engineering, and construction (AEC) industry is responsible for roughly 13% of global GDP. It designs and builds the hospitals we're born in, the offices we work in, and the infrastructure that connects everything in between. And yet, it remains one of the least digitized industries on the planet.
That's starting to change — fast.
The Data Problem Nobody Talks About
A single mid-size commercial building generates thousands of files during its lifecycle: architectural drawings, structural calculations, MEP layouts, energy models, compliance documents, cost estimates, and scheduling data. Multiply that across a portfolio of dozens or hundreds of buildings, and you're looking at terabytes of deeply technical, highly structured data.
Here's the problem: almost none of it talks to each other.
IFC files sit in one system. Cost data lives in spreadsheets. Compliance checks happen manually, often by consultants reviewing 2D prints with a highlighter. The industry has digitized its documents without digitizing its workflows.
This is exactly the kind of problem AI was built to solve.
From Document Intelligence to Building Intelligence
When most people hear "AI in construction," they think of generative design tools or robots laying bricks. Those are real, but they're not where the biggest impact is happening today.
The real transformation is in building data intelligence — the ability to take raw, complex building information and turn it into something queryable, validatable, and actionable.
Consider what becomes possible when AI agents can natively understand an IFC model:
- Automated code compliance. Instead of a human cross-referencing fire codes against floor plans, an AI agent can validate an entire model against multiple jurisdictions in minutes. Not by pattern-matching keywords, but by understanding spatial relationships — exit distances, compartment boundaries, material fire ratings.
- Instant cost estimation. Traditional quantity takeoff requires a specialist to manually measure elements from drawings. An AI agent that understands building geometry can extract quantities, map them to cost databases, and produce estimates that would take a human team days to assemble.
- Natural-language queries. Imagine asking "What's the total glazing area on the south facade?" or "Show me all rooms that don't meet accessibility clearance requirements" and getting an answer in seconds, directly from the model. No specialized software training required.
- Continuous validation. Rather than checking compliance at milestone reviews, AI agents can validate models continuously as they evolve — catching issues when they're cheap to fix, not after they've been built.
Why Now?
Three things have converged to make this moment different from previous waves of "construction tech" hype:
1. Foundation models understand structure. Large language models have gotten remarkably good at reasoning over structured and semi-structured data. When combined with domain-specific training on building standards, material properties, and spatial logic, they become genuinely useful reasoning engines — not just text generators.
2. IFC is maturing as a standard. The Industry Foundation Classes specification has been around for decades, but adoption has reached a tipping point. With governments mandating BIM deliverables in IFC format (the UK, EU, Canada, Singapore, and others), there's finally a critical mass of standardized building data to work with.
3. Graph-based AI is production-ready. Buildings are inherently graph structures — spaces connected to spaces, systems routed through zones, components assembled into assemblies. Graph neural networks and knowledge graphs are now mature enough to model these relationships at scale, enabling reasoning that flat data formats can't support.
What's Holding the Industry Back
If the technology is ready, why aren't more firms using it?
Integration complexity. Most AEC firms run dozens of software tools that were never designed to work together. Getting data out of Revit, into a graph database, validated against a code corpus, and back into a BIM environment requires deep technical plumbing that most firms don't have in-house.
Trust deficit. Engineers and architects are personally liable for their designs. They need AI that can explain its reasoning, cite the specific code clause it's checking against, and show its work. Black-box predictions aren't enough.
Talent gap. The intersection of building science and machine learning is a very small Venn diagram. Most AI teams don't understand IFC semantics, and most AEC teams don't understand transformer architectures. Bridging that gap requires a new kind of company.
The Agent Architecture Approach
At Intento Labs, we believe the answer isn't a single monolithic AI model — it's a constellation of specialized agents, each expert in a specific domain, working together.
An IFC Parser agent that understands building geometry. A Fire Code agent trained on NBC, OBC, IBC, and Eurocode requirements. An MEP agent that can trace duct routes and pipe networks. A Cost Engine that maps elements to RSMeans data. A Scheduler that understands construction sequencing.
Each agent is independently testable, auditable, and updatable. When a building code changes, you update one agent — not retrain an entire system. When a client has a domain-specific requirement, you add a new agent to the constellation.
This modular approach means the system gets smarter with every deployment, every edge case, every new jurisdiction — without becoming more fragile.
What Comes Next
We're still in the early chapters of this transformation. The firms that move first won't just be more efficient — they'll be able to offer services that are literally impossible today: real-time compliance monitoring, automated permit-ready documentation, predictive maintenance from design data, and natural-language interfaces that make building intelligence accessible to everyone, not just BIM specialists.
The AEC industry doesn't need to be disrupted. It needs to be unlocked. The data is already there. The standards are maturing. The AI is ready.
The only question is who moves first.
---
*Intento Labs builds AI agents that transform building data into actionable intelligence. We're currently in early access with select partners across North America. If you're working on complex building data challenges, we'd love to hear from you — reach out at hamid@intentolabs.io.*