Design Intent Is What Matters

A mid-size architecture firm recently posted for a "Design Quality Assurance Manager." The job listing read: review design changes against project intent. Catch scope creep before it costs money. Validate that modifications align with owner requirements.

That role exists because firms are drowning in design change management. Proposed changes come in constantly. Someone has to understand what the design actually is, and what the design should be, to catch the misalignment between them.

The pattern is bigger than one job posting. AI in AEC is moving from "answer questions about the model" — search, summarize, retrieve — to "understand intent and validate against it." Modify. Optimize. Catch errors before they ship. The firms that can encode design intent into their workflows will have AI that doesn't just analyze — it can participate in decision-making.

That's why the system underneath matters so much. You can't teach AI to understand design intent if your data doesn't express intent. You need a system that says: this wall is structural load-bearing. This opening is required egress. This dimension is tied to code.

The tools to start exist now. Claude with MCP — Model Context Protocol, a way to give AI structured access to your existing tools — can connect to Revit and read drawings. But the real work isn't installing tools. It's teaching the system what "intent" means in your projects specifically. That requires a model of how your design decisions actually work.

If you're evaluating AI tools for your firm, the question worth asking is: can it understand intent, or just summarize what exists? The summarization tools are nice-to-have. The intent tools are competitive advantage. And you can't get those unless you've modeled your own design logic first.

Start with one workflow: design change validation. Document what makes a change valid or invalid in your projects. That's your intent model. AI learns from it.