You’ve received the proposals. They’re all professional, well-organized, and full of confident claims about team experience, project approach, and commitment to your success. And you have no idea which one is actually telling you the truth.

That’s always been the owner’s problem with AEC proposals. AI is about to make it more complicated, and also, potentially, more revealing.

Proposals are getting faster to produce

AI tools can now do in hours what used to take a proposal team days. Pull relevant past project descriptions from a content library, adapt the standard approach narrative to match the specific project requirements, format the fee schedule, generate a cover letter that sounds personal and specific.

The work product looks the same. It might actually be better, more consistent, fewer typos, tighter formatting. But it took a fraction of the time and a fraction of the skilled attention.

That’s not inherently bad. If AI is freeing up senior people from the production work of proposals, they should be thinking harder about strategy, what this client actually needs, what approach would genuinely differentiate their offer, where the project risks are and how to address them.

The question is whether that’s actually happening, or whether the time savings from AI-assisted production are just going to more proposals, more shortcuts, and less substantive thinking.

What to look for in a well-prepared proposal

The signal in a proposal has never been the formatting or the photography. It’s whether the submitting firm has thought seriously about your specific situation.

Does the proposal demonstrate understanding of your business, not just your building program? Does it identify the specific risks on your project and propose a credible response? Does the team composition match the actual challenges, rather than showcasing whoever the firm considers its best credential for this market?

When AI helps with proposal production, it doesn’t change what the right answers to those questions are. It just changes how easy it is to produce words that sound like the right answers.

The counterfeit proposal problem, submissions that look sophisticated but represent no real thinking, is getting harder to detect. But the counterfeit-detection approach hasn’t changed: ask specific questions, in person, about the specific situation, and see whether the people across the table can answer from genuine knowledge.

The proposal interview as the real evaluation

If your selection process allows for interviews, use them properly. Not as a formality after you’ve already decided based on the written submission. As the primary evaluation.

Ask the PM who will actually run your project, not the principal who will introduce them and disappear, to walk you through how they’d handle a specific problem type on your project. Not a hypothetical. An actual scenario drawn from your project’s real constraints and risks.

Ask them what questions they’d want answered before starting design. If they can’t generate substantive questions, they haven’t thought about your project.

Ask what they’d do if the budget showed a significant gap at the schematic design phase. The answer will tell you whether they’ve developed a genuine approach or whether they’re going to wing it and hope.

What AI-powered proposals mean for your evaluation criteria

The practical upshot: written proposals matter less than they used to, because production quality has been commoditized. The information they carry is still valuable, but you should weight it lower and the interview conversation higher.

The firm that produces the slickest submission isn’t necessarily the one that has thought most clearly about your project. And the firm whose proposal looks a little rougher around the edges might have poured more thinking into the substance.

Evaluate for judgment, not production value. That’s always been the right approach. AI just makes it more urgent.