The contractors getting traction with AI aren't trying to transform everything at once. They're picking one problem, proving it works, and letting success build momentum.

In conversations I’ve had with IT leaders, founders, and builders - it can be hard not to experience shiny object syndrome. With a steady stream of tools that promise to 10x your business overnight, how do businesses know where to begin? It feels like a constant ebb and flow in the construction tech industry between overwhelming enthusiasm for new tools and overwhelming confusion about deployment and governance. Where does “the AI live”? What if it hallucinates? How will I know if it’s worth the investment? My advice is to start simple with writing tools, calculators, and regulation lookups to establish the platform, delivery method, and security before tackling bigger projects.

A few principles that seem to hold:

The risk of skipping this step is what one client called "work slop" - people using AI without understanding the outputs and passing poor quality work forward. AI can compound issues faster if not used thoughtfully, which makes expert oversight crucial.

There's another reason to start small that's easy to overlook: it gives you room to learn and correct before the stakes get high. Starting small is not only about proving value but also having the runway to make mistakes while they are cheap.

A question came up in a client meeting recently that captures this idea well. The controller asked: "Are we buying a Lamborghini or a Toyota?" My answer was Toyota. Reliable, cost-effective, and you're less likely to crash at high speed. There is so much hype these days that makes businesses feel like they need to throw AI at every problem. In reality, we should be throwing focused conversations, domain expertise, and planning at the problems - and yes, AI might be a useful tool along the way.

One of the conversations I mentioned above revealed another common concern for construction businesses. How do we evaluate when to create a tool from scratch versus pay for an existing solution?

This is the build vs. buy question that a lot of folks are considering. Here's how I think about it:

Buy when:

Build when:

Most contractors I talk to are somewhere in the middle. They're using platforms like Procore or BuilderTrend for project management but have specific pain points those tools don't address. Things like voice-to-documentation for field workers, quick regulation lookups, or calculators.

I think a hybrid approach is what most businesses will need. Use platforms where they're strong. Build custom tools for the gaps. The key is being honest about which problems are truly unique to your operation and which ones are just unfamiliar or underutilized.

A December report from Dodge Construction Network found that 87% of contractors believe AI will have a meaningful impact on their business. That's a striking number for an industry often characterized as slow to adopt technology.

But the more interesting finding is what's holding people back. Only 26% of contractors rated their current data quality as high. Data accuracy and security were the top concerns, followed by implementation costs and internal resistance.

The optimism is real - contractors know AI is here and most believe it can help their business. But the gap between believing and doing remains wide. The report found that only about 40% have dedicated budget to AI, and 38% are creating implementation teams.

The takeaway for me is that 2026 is going to be less about new tools and more about the hard work of getting existing tools adopted. The technology isn't the bottleneck. Data quality and change management are.

One of the most common (and most valuable) use cases for AI is using it as a thought partner. So I built something in Claude that's been incredibly useful.

It's called an advisory board skill. You pick thought leaders whose perspectives you value and teach Claude about their philosophies. Then you can consult them on business decisions.

I use mine to review client opportunities, think through strategic moves, and check my assumptions. It's not a replacement for real advisors or your own judgment, but it dramatically improves the output when you need to look at a problem from multiple angles.

The setup takes about 20 minutes if you already use Claude. I put together a guide if you want to build your own:

Advisory Board Skill Setup Guide