Ask your architect or owner’s rep where they’re using AI. Most will describe things like drafting proposal sections, summarizing contracts, routing invoices faster, cutting meeting prep time.

Real efficiencies. The wrong target.

Where the money is made and lost

In an AEC firm, money moves upward from projects. Revenue enters at the project level, pays direct labor and expenses first, and whatever’s left flows up to fund everything the firm runs that clients don’t pay for directly: administration, marketing, accounting, leadership overhead.

Overhead is necessary. But it’s funded entirely by project margin. The firm only earns what it can skim off the top of what projects produce.

Overhead, for the most part, consists of highly repetitive, highly structured work. Invoice routing, timesheet processing, proposal assembly, onboarding paperwork. That’s not a frontier AI problem. It’s a process problem. Good process mapping and a capable systems integrator solve most of it without AI anywhere in sight.

When firms use AI to cut 20 percent of their overhead, they improve their total cost structure by maybe seven points. Meaningful. Finite. And it doesn’t touch the thing that actually generates margin.

Where AI should actually go

Every project is unique. Different site, different owner, different team, different code context, different buried assumptions. No two are the same.

That’s exactly the kind of problem AI was built for.

The veteran project manager is valuable for one reason: thirty years of pattern recognition built across jobs that were all different. They’ve seen this coordination problem before. They know which submittal always runs late. They can sense when the schedule is drifting before the report catches up. That expertise is built from novelty, from a career of situations where no fixed rulebook applied.

AI excels in exactly that domain. Not processing thousands of identical invoices. Processing a unique, time-bound web of decisions (this project’s drawings, this project’s contracts, this project’s RFI history) and surfacing the patterns inside it.

The speed a design team resolves a coordination conflict before it becomes an RFI. Whether a submittal clears in two days or twelve. How many projects one PM can carry before things fall apart. The soft scope assumption that quietly becomes a change order.

Every one of those feeds project margin or leaks it. AI pointed at projects touches all of them. AI pointed at overhead touches none of them.

What this means for how you evaluate your team

When your owner’s rep or contractor tells you they’re using AI to serve your project, ask specifically what it’s doing.

If the answer is mostly back-office (faster reporting, automated documentation, streamlined communication) that’s efficiency that benefits their operations. Your project gets the same service, packaged more efficiently.

The answer that changes your outcomes: AI that monitors your project’s specific data in real time. That flags a submittal delay before it cascades into a procurement problem. That surfaces a coordination conflict between disciplines before it becomes a field problem. That tracks budget trajectory against early-warning indicators rather than waiting for the monthly report.

That’s AI pointed at the project. That’s where the margin lives. That’s what changes what you pay and when you move in.

The practical question

Your next project team will likely tell you they use AI. The right follow-up: show me how it changes what happens on my project, not what happens in your office.

One is a cost story for them. The other is a value story for you.