Yesterday I gave a keynote to a room of about a hundred construction leaders. Subs, GCs, trades. People from all over the country, running very different kinds of businesses, sitting in the same room for the same reason.
Q: Is this saying most construction companies shouldn’t use AI at all? No. It’s saying most companies shouldn’t try to build AI. Using well-built tools is completely different from hiring engineers and building your own. Almost everyone in construction should be using AI tools. Almost nobody should be building them from scratch.
Q: How do I know if I’m a “self-perform” company or not? Ask the same questions you’d ask about self-performing any trade. Do you have the volume to justify the investment? Do you have the balance sheet to absorb a few years of cost before payoff? Do you have the patience to stick with it when it’s slow and expensive in year one? If you’re hesitating on any of those, you’re not a self-perform company. That’s fine. Most aren’t.
Q: What size company can actually justify building internal AI capability? There’s no clean revenue cutoff, but directionally: this is a game for GCs doing billions in annual volume, with the margin and cash position to fund a multi-year engineering investment without it disrupting the core business. For everyone else, the math on hiring expensive software talent rarely clears.
Q: If I’m not building AI in-house, what should I actually be doing? Get good at choosing tools. That’s the real skill. Figure out which vendors solve real problems on your jobs, pilot a few, hold them accountable, and integrate the ones that work into how your teams already operate. It’s the same discipline you use to vet and manage a good sub.
Q: Isn’t it risky to fall behind if I don’t invest in AI now? The bigger risk is spending a year and a real budget building something in-house that a vendor already built better and cheaper. That’s not falling behind. That’s an expensive detour. Falling behind actually looks like ignoring good tools that already exist, not skipping the engineering build.
Q: Does this apply to subs and trades the same way it applies to GCs? Yes, maybe even more so. Subs and trades almost never have the volume or balance sheet to justify a self-perform AI bet. The subs who’ll win are the ones who get sharp about picking two or three tools that genuinely speed up estimating, tracking, or scheduling, not the ones trying to build something themselves.
Q: Where did this idea come from? From listening, not lecturing. I gave a keynote to about a hundred construction leaders, subs, GCs, trades from all over the country, and this was the pattern that kept surfacing in the questions and conversations afterward. The room already sensed the tension between “we need to do AI” and “we don’t have the resources to do AI the way the big guys do it.” This is the reframe that resolved it.
Q: What’s the one thing you’d want a reader to walk away with? There isn’t one right way to win with AI in construction. There’s the way that fits your business, and the way that doesn’t. Knowing which one you’re in is more valuable than chasing a strategy built for a company ten times your size.
