Almost every AEC firm has ChatGPT or Microsoft Copilot deployed now. Almost none of them have done anything meaningful to train their teams to use it.

That’s not a technology problem. It’s a management problem. And the productivity it’s costing you is very real.

The “we have the tools” trap

Buying software and deploying software are different things. Every project manager in the world has access to email. That doesn’t mean every project manager communicates effectively.

When a firm subscribes to an AI platform and sends their team a Slack message saying “here’s your login,” they haven’t adopted AI. They’ve licensed it. Those are different things.

The gap between licensing and adoption shows up in how people actually use the tool. Most teams default to the most obvious use cases: ask ChatGPT to write an email, clean up some text, summarize a document. Useful, but it barely scratches the surface of what’s possible.

The genuinely transformative uses, systematically extracting risks from contract documents, synthesizing project status from multiple data sources, generating first-pass specifications from scope briefs, cross-referencing submittals against specifications. Those require workflow design, not just tool access.

The 12-minute rule

Here’s a practical frame that makes this concrete. If every person on your project team found 12 minutes of daily work that AI could accelerate, you’d capture roughly an hour per person per week. Across a 20-person project team, that’s 20 hours per week, the equivalent of a half-time employee, in productivity that can go toward higher-value work.

But that only happens if people know what 12 minutes looks like. What tasks are good fits for AI acceleration? How do you prompt effectively to get useful output? What do you do when the output is wrong or incomplete?

None of that is intuitive. It requires deliberate training that most firms aren’t doing.

What real training looks like

There are two models for building AI capability in a firm. One works. One doesn’t.

The model that doesn’t work: send everyone to a general-purpose AI training. Spend two hours on prompting techniques. Remind people the tools exist. Move on. Six weeks later, nothing has changed.

The model that works: train the trainers. Identify two or three people who are genuinely curious about AI and willing to invest in becoming fluent. Give them time and latitude to experiment. Task them with developing specific AI workflows for specific tasks your firm does repeatedly. Then deploy those workflows across the team with targeted training on exactly those use cases.

The second model requires more upfront investment and produces dramatically better outcomes because it generates workflows tailored to how your firm actually works, with champions who can support adoption and troubleshoot problems.

The policy problem that goes largely unaddressed

Most firms that have AI policies have AI policies that say what you can’t do, don’t share confidential client information, don’t rely on AI for factual claims without verification, don’t use outputs without review.

What most AI policies don’t address: what you should do. What workflows benefit from AI assistance. What the quality bar is for AI-assisted work product. How to document when AI was used and how output was reviewed.

The absence of positive guidance means individuals are making their own judgments about appropriate use, producing wildly inconsistent results across the firm. One project manager is using AI to dramatically accelerate their work. Another on the same project is doing everything manually because they’re not sure what’s acceptable.

That inconsistency is a quality problem as much as a productivity problem.

What this means for owners

As a building owner evaluating the firms you hire, this dimension of capability is worth probing. Not whether they have AI tools, everyone has AI tools. Whether they’ve invested in knowing how to use them.

Ask specifically: what AI-assisted workflows have you deployed for project management, and what’s the adoption rate among your project staff? Ask for a concrete example of how AI changed the way they worked on a recent project.

The firms that can answer these questions specifically have done the work. The ones that give you vague assurances about their AI investments haven’t.

AI fluency at the firm level translates to real outcomes at the project level: faster deliverables, more thorough document review, earlier risk identification, better information for your decision-making. The gap between firms that have built that capability and firms that haven’t is widening.

The tools are everywhere. The training is not. That’s where the competitive advantage lives for the next few years.