Walk into most contracting firms right now and you’ll find AI already in the building. Someone in the office is using it to draft proposals. An estimator is running it against a spec sheet. A PM is using it to summarize a stack of RFIs before a Friday call. None of this happened because leadership rolled out a strategy. It happened because individual people got tired of doing slow work slowly.
That’s the real state of AI in construction and specialty contracting today. Everyone’s touching it. Almost no one’s running it.
It shows up as a collection of personal habits rather than a business capability. One estimator has a workflow. The PM next to them has a different one, or none at all. Nobody’s comparing notes, because nobody was ever asked to. The firm ends up with pockets of speed that never connect to each other, so none of it compounds. It just sits in individual inboxes, doing individual favors for individual people.
That’s not a knock on the people doing it. If anything, it says the workforce is more resourceful than most leadership teams give it credit for. The problem is what happens next, or rather, what doesn’t.
A tool is not a system
A single person getting faster at a task is a convenience. A firm getting faster at winning work, controlling costs, and protecting margin is a strategy. Most contractors are living entirely on the convenience side of that line right now, and don’t realize it.
Getting to the other side takes three things individual tool use never produces on its own. A shared understanding across the team of what AI is actually for inside the business. A small number of use cases the firm chose deliberately, instead of ones it backed into. And enough oversight that the work being sped up is still getting checked by someone who understands the job.
None of that requires a software rollout or a six-figure platform commitment. It requires a decision. Most firms haven’t made one yet.
Start with the problem, not the tool
Every contractor evaluating AI right now is answering the wrong first question. The question isn’t “which tool.” It’s “which problem.”
A firm bleeding hours in estimating has a different starting point than one losing bids because its past work never turns into a story anyone can use in a proposal. A firm drowning in change order paperwork has a different starting point than one that loses institutional knowledge every time a senior person retires. Point AI at the wrong problem and even a genuinely capable tool underdelivers, not because the technology failed, but because nobody aimed it. Point it at the right one, and a fairly modest use of AI moves a real number: fewer hours per bid, faster closeout, a proposal that actually reflects the strength of the firm’s work.
That’s the shift worth watching over the next year. Not whether contractors are using AI. Nearly all of them, in some form, already are. It’s which firms turn scattered individual habits into something the business runs on, and which ones stay exactly where they are now: busier, faster in a few corners, and no further ahead than when they started.
The firms pulling away aren’t doing anything exotic. They’re choosing on purpose instead of accumulating by accident. That’s a smaller leap than it sounds, and it’s just as available to a 30-person mechanical contractor as it is to a national GC.