A preconstruction lead at a general contractor told me last month that he’d stopped counting takeoff software demos. Not because the demos were bad. Because after the seventh one, he realized he was watching the same product wearing different colors.
“I don’t need a faster way to click on drawings,” he said. “I need my best estimator to stop clicking on drawings.”
I thought about that for a long time after the call.
For thirty years, construction software has been sold as a better tool. Better takeoff. Better clash detection. Better bid leveling. Each one promising to shave hours off a specific task. And each one, at the moment of adoption, handed back to the same estimator who was already working late.
The tool got faster. The estimator did not get her evenings back.
The pitch
When we started Boon, we thought we were building a better takeoff tool too. The pitch was clean. Estimators spend twenty hours a week measuring things on drawings. If we could get that down to two, the ROI would sell itself.
The customers we won early proved the pitch was right on the math. A national electrical contractor cut a specific class of takeoff from eighteen hours to under two. The numbers were real. We published them.
But the same customers kept asking for things we hadn’t planned to build.
Could Boon read the specifications and flag the places where the drawings disagreed? Could it draft the RFI to the design team when it found a conflict? Could it look at four subcontractor bids and tell them which one was actually the cheapest once you accounted for everything each one had quietly excluded?
These weren’t takeoff features. They were the parts of the job that the tool was never meant to touch. The parts the estimator did after the takeoff was done, at 9pm, from memory, while eating cold food.
We were not building a better takeoff tool. We were building the coworker who could do the rest of the work.
The real category
There’s a pattern in enterprise software where the incumbent tools optimize the fast part of a job and leave the slow part alone. The slow part is where the human still is.
Estimating is a good example. The mechanical act of measuring quantities off a drawing is roughly fifteen percent of what a preconstruction team actually does. The other eighty-five percent is the reading, the reconciling, the writing, the second-guessing, the going back to the architect for the third time to ask what they meant by “as required.” No takeoff tool has ever tried to help with any of that, because a takeoff tool is a takeoff tool. It has a shape. The shape doesn’t fit around a job.
The category isn’t “AI takeoff software.” The category is “the person who does the rest of the work.”
We call that person the Boon Agent. It reads the full plan set the way an estimator would read it. It notices the places where the drawings and specs contradict each other. It writes the RFIs. It compares the sub bids on a real like-for-like basis. It answers the project manager’s questions about which addendum changed what. It doesn’t ask you to open a separate application or learn a new interface. It works inside Slack and Teams and the tools your team already uses.
It’s not a takeoff tool. It’s a teammate, and takeoff is one of the things it can do.
Why now
The construction industry has been sold “better tools” for so long that the phrase has stopped meaning anything. Every conference has a booth with the same three claims about accuracy, speed, and ease of use. Every demo ends with a pricing page.
What’s changed in the last eighteen months is not the tools. It’s what the underlying technology can do. Foundation models can read a two-hundred-page specification and hold the whole thing in their head. They can cross-reference it against a drawing set. They can write. They can compare. They can notice when something on page 47 contradicts something on page 183. None of that was possible in a takeoff-tool era, and none of it is really about takeoff.
The firms that will win the next decade of preconstruction are not the ones that bought the faster takeoff. They’re the ones that gave their best estimators a teammate who could carry the eighty-five percent that no tool has ever tried to help with. The estimator does the judgment work. The teammate does the reading and the writing and the checking. The estimator gets her evenings back, and the firm bids more work with the same people.
That is a different product than a takeoff tool. It has a different shape. It’s sold differently, it’s priced differently, and it changes what an estimating team is actually for. It’s also harder to build, because the surface area is larger and the failure modes are less forgiving. A takeoff tool can be wrong about a length. A teammate can be wrong about what the drawing says the contract includes. The bar for the second one is a lot higher, which is a good problem to have.
The site rewrite
Reader beware: this is the part where the person who runs the company tells you what they did. Bear with me.
The homepage of getboon.ai used to talk about AI agents that automate takeoff, estimating, and bidding. All three of those words are still true. But leading with them told the wrong story. It made us sound like the eighth booth at the conference.
The new line, above the headline, is quiet and specific:
Your AI teammate for preconstruction.
The rest of the site is being rewritten to match. Not because we’ve discovered a new marketing hack, but because we finally have language for what we’ve actually been building for the last three years. The takeoff numbers are still there. They’re just no longer the point. The point is what the estimator does with the hours the takeoff no longer takes.
Still probably wrong
I don’t think we have this fully figured out. For a long time, we described Boon Agent in terms of the tasks it does, because tasks are easier to demo than a teammate is. That was probably a mistake, and it made the product feel smaller than it is. Every one of our best customers uses the agent for something we didn’t put on the pricing page.
We will get this framing partially wrong again. The category is new, the language is still forming, and the customers who buy first are the ones who already had a sense of what they were missing. The customers who buy next will make us say it more clearly.
What I keep coming back to is the preconstruction lead’s line about wanting his best estimator to stop clicking on drawings. He wasn’t asking for a faster clicker. He was asking for the rest of the job to get easier, so the person he pays the most for her judgment could spend her time on judgment. That’s the product. It has always been the product. We just kept selling the clicker because the clicker was easy to explain.
If you’re building a preconstruction team right now, the useful question is not which takeoff tool to buy. It’s which teammate to hire. There aren’t very many yet. The ones that exist will define what the next generation of estimating actually looks like.
