By Mark Schwartz, President, AECO, Trimble
A non-renewable resource
I like to call AI a force multiplier. It's a military term that refers to anything that makes a unit far more effective than its size alone would suggest, and that is what good AI does for a construction operation.
There is a second phrase from the same world I think about just as much. The clean version is "one aw-crap destroys a thousand atta-boys." Anyone who has run a jobsite knows that math. You can do everything right for months and lose all of it in a single afternoon. Trust works the same way. It builds slowly and it spends fast, and once it is gone you do not get it back on the schedule you would like, if you can even get it back at all.
Put the two phrases side by side and you have the whole problem. AI gives you leverage you have never had. A jobsite gives you no room to be wrong at that scale. So when a project manager hesitates to turn AI loose on a live site, that is not timidity.
AI gives you the leverage to scale, but trust is the foundation you need to bear the load.
Higher stakes
Most software lives in a forgiving world. An AI agent writes a bad caption or misroutes a support ticket, and the cost is an afternoon and an apology. Construction does not get that margin. The same wrong answer here lands on a pour schedule, a payment run or a crew standing on a deck waiting for a number.
It also changes what you can accept from the people selling you the tool. A vendor moving into a forgiving industry can ship fast and clean it up later. A vendor moving into yours cannot, and the good ones already know it.
Looking past the hype
Hype has flooded the market. It’s exhausting—I’m exhausted by it. So it matters more than ever to sort the companies that are there to help you do the work from the ones that are there to close the sale and move on. The good ones share three traits, and you can check for all three before you sign anything.
They treat your data as yours: The information a business hands an AI system belongs to that business. A trustworthy company keeps that line in plain sight, because it has nothing to hide on either side of it.
They show their work: You can trace what happened from the input all the way to the result. Nothing important is a black box. Trust is auditable. It leaves a record.
They demonstrate accountability: Real accountability is not a good apology after something breaks. It is proactive—catching the problem before it reaches you, fixing the cause instead of the symptom, and owning the long-term fix rather than the quick patch.
How do you know if it’s the right AI tool for you?
In my experience, the AI that matters in this industry is human-aided AI. Today, AI is not going to hammer a nail into a wall. What it can do is put the crew in the right place at the right time so they can hammer once, hammer fast, and know where to go next.
So the first question to ask about any AI tool is: Does this match a workflow I actually run, with a role I actually have on my team? If the answer is yes, the tool can be evaluated. If the answer is some version of "well, eventually," that's a sign the tool is being sold ahead of where the industry can use it.
The second question is whether you can verify good results. Start with a workflow your team knows in its sleep. Submittal reviews, daily reports, takeoff estimates. Familiarity allows you to tell when the AI is doing the work well. The closer the first use case sits to existing muscle memory, the faster trust is built.
The third question is what the tool doesn't do. The most trustworthy AI pitches I've seen explicitly name the boundaries. They tell you which workflows are ready and which aren't yet. They tell you what the tool can't catch and what still needs a human eye.
When you know the work and you know the technology’s limitations, you know what you can actually trust it to do.
How do you implement it into your workflow without disruption?
Document your workflows. Know what each step does, who owns it, what tools are involved, where the data lives. The vendor is fitting their tool to your process. You should know that process better than they do. If you can describe the workflow in writing well enough that a new hire could follow it, you're ready. If you can't, that's where to start.
The tool itself gets you the first jump in productivity. Ten percent, twenty percent, maybe more. That's real. But the next twenty or thirty percent above that comes from restructuring the work around the tool. The workflow you documented in step one is going to change once AI is part of it. The companies that get the second jump are the ones that plan that restructuring as a separate step, scheduled in advance, with the same intentionality they brought to the first deployment.
The last move is to expect that you'll fail at some point, and by “fail” I mean learn. You will. The tool will return the wrong answer, or miss something obvious, or do something nobody predicted. The teams that adapt the fastest are the ones that build a review cadence into the deployment from day one. Every few weeks, the people using the tool sit down and talk about what worked, what surprised them and what they want to bring up with the vendor next. The conversations get sharper over time. So does the system. So does the partnership.
Trusting tomorrow
The trust standards being built right now are getting easier to ask for. Identity controls, audit trails and accountable response paths were enterprise-only conversations a few years ago. Today they are becoming the floor for the whole industry, regardless of company size.
Trust culture compounds the same way safety culture does. One generation of crews learns it, and the next generation grows up assuming it. The same arc is starting to take shape for AI. The crews everyone wants to work with five years from now are the ones building the right habits today.
Hear more from Mark Schwartz and leaders from Microsoft, NVIDIA and Snowflake in the “Beyond the Hype” AI webinar recording.
About the author
Mark Schwartz serves as President, AECO at Trimble, where he oversees construction enterprise solutions, civil infrastructure software, and owner and public sector businesses. A champion of digital transformation (Dx), Schwartz previously served as Trimble’s Chief Digital Officer, leading the modernization of the company's enterprise systems and infrastructure to spearhead its transition to an "as-a-service" business model. Today, he continues to push the envelope of industry innovation by overseeing the development of Trimble's Agentic AI Platform, delivering purpose-built, productivity-driven AI solutions to physical industries.
