Every architecture and engineering firm in your inbox will tell you they use AI.

The question is whether any of it shows up on your project.

The procurement framework that’s broken

For decades, building owners have selected design firms the same way. Review the portfolio. Check references. Evaluate qualifications. Negotiate the fee.

This worked reasonably well when the core capabilities of firms were relatively stable. A firm that was excellent at hospital design five years ago was probably still excellent at it today. The skills were the same. The tools were different but not transformatively so.

That’s changing fast. AI is altering what design and engineering work actually looks like: how coordination happens, how documentation gets produced, how RFIs get resolved, how closeout packages get assembled. The firms that have genuinely integrated AI into their workflows are becoming measurably more capable than the firms that haven’t.

And the standard procurement evaluation doesn’t distinguish between them.

You can evaluate past work. You can check references. But none of that tells you whether a firm’s AI adoption is substantive or performative, whether it will change your project outcomes or just their marketing materials.

What the research actually shows

The construction industry wastes roughly 40 cents of every capital dollar on inefficiencies embedded in the delivery chain. That number has been remarkably consistent for decades. Fragmented teams, incomplete information transfer between phases, gaps between design intent and field execution. These patterns repeat endlessly.

AI has the potential to address many of these root causes. It can catch coordination conflicts before they reach the field. It can surface RFI patterns that indicate underlying design gaps. It can compress the time between a problem appearing and a decision being made.

But only in the hands of firms that have actually built it into their workflow. A firm with a ChatGPT subscription is not the same as a firm that has redesigned its production process around continuous AI-assisted coordination.

The questions worth asking

Sophisticated owners are starting to add AI evaluation to their firm selection process. Not as a checklist, but as a genuine capability assessment.

Show me the RFI data. Ask for the RFI volume and resolution time on the firm’s last three comparable projects. If AI is genuinely in the coordination workflow, these numbers should have changed in the last two years. Faster resolution, fewer field RFIs, better documentation. If the numbers are the same, the AI adoption is cosmetic.

Walk me through a specific problem AI caught. Not a category of problems, a specific instance. A coordination conflict between structural and MEP that showed up in model review before it reached the field. A specification gap that a check caught three weeks into design development. Substantive answers are specific. Performative ones are general.

What does your closeout package look like? If a firm’s AI workflow is generating better documentation during design and construction, the handover package should reflect it. Queryable data, organized equipment records, complete warranty documentation. If the answer is still “we’ll give you the PDFs,” the AI isn’t reaching the output that matters most to you.

Who owns this internally? A firm serious about AI has someone accountable for it with a real budget. Not a committee studying it, not a pilot project that’s been “ongoing” for eighteen months. An actual function with ownership and results.

The emerging standard

A group of institutional owners developed a structured assessment framework for exactly this purpose, a systematic way to evaluate how design and engineering firms actually deploy AI, and whether that deployment produces better outcomes for the building owner.

The assessment covers several dimensions: whether AI is embedded in core production workflows or siloed in a separate team; whether the technology genuinely catches coordination issues earlier; whether the documentation it produces is actually useful for operations; and whether the firm has invested in training its people to use these tools effectively.

It’s the first time owners have collaborated to create a shared evaluation standard for something this specific. That it happened reflects how seriously sophisticated owners are taking this question.

What you should do before your next selection

Before your next design firm shortlist, add an AI capability evaluation to your process. Not a question about whether they use it, every firm will say yes, but a structured assessment of how it shows up in actual project outcomes.

The firms doing real work here will welcome the scrutiny. The ones engaged in theater will hedge and generalize.

That difference tells you more than the portfolio.