Ask a senior estimator how they priced a job and you'll usually get two answers.
The first is the one in the file: the takeoff, the spreadsheet, the quantities pulled off the drawings.
The second takes longer to come out because they never wrote it down. The engineer on this job always undersizes the risers, so you check the shafts before pricing the sleeves. This GC's addenda land two days before bid day, so you hold a little back until you see them. None of that is in the file, but all of it is in the number.
Think of them as two models of the same project. One lives in software. The other lives in someone's head. For 20 years, the industry has poured its money into the first one, betting that good enough software would make the second matter less.
That bet was only half right. Software has made jobs faster to measure and easier to coordinate. It hasn't made judgment any less important. If anything, the more detailed the software gets, the more a job depends on someone who knows how to read it.
Where the money leaks
A takeoff can get every count right and still cost you money. The drawings won't tell you the specified chiller is 30 weeks out or that the only site entrance can't take a concrete truck. The estimator who's been burned before will. Nobody else knows, because that lesson lives in an old email, a text, or a phone call. When that estimator leaves, the lesson goes with them.
It doesn't take a big miss to hurt. According to CFMA's 2025 Financial Benchmarker, the typical contractor's pretax margin was 6.7% in 2024. At that rate, $10 million of revenue returns about $670,000. A single 2% miss can take $200,000 of it, nearly a third. That's what one experienced estimator's judgment is quietly protecting.
Experienced people are short on time, not knowledge
Most estimating teams don't lack judgment. They lack time. Their best estimators spend much of the week measuring: counting fixtures, tracing duct runs, and working out square footage. Nobody hires a veteran estimator to count sprinkler heads, yet that's where a lot of their week goes.
Give that time back, and it can go where it matters: calling the supplier, reading the spec twice, and asking the scope question nobody else thought of.
What this means for how tools get built
AI is moving into estimating fast. In the 2026 Construction Hiring and Business Outlook from AGC and Sage, 61% of firms said they were using AI or planning to invest more in it. After office and administrative work, estimating and preconstruction were the most common uses.
So what should the software do? The counting. That's where it's fast, and people are slow. Scope, risk, and relationships belong with the people who understand them.
That's the idea behind Beam AI, our takeoff platform. It reads full plan sets and specifications and returns quantities across trades, instantly on a self-serve takeoff or, for complex bids, after our QA team reviews the AI's work. We're not trying to hand anyone a finished bid. The aim is a takeoff they can trust, so their hours go into the calls only they can make.
JBG Flooring and Installation is a good example. Manual takeoffs used to eat so much of the week that the team was missing projects. Now they save up to two days per takeoff. That time goes into better supplier pricing, confirming materials are available, and making sure each bid is competitive.
The knowledge is walking out the door
Here's the harder problem: Many of the estimators who carry the second model are close to retiring. Associated Builders and Contractors estimates the industry needs 349,000 new workers in 2026, mostly to replace retirees. In the 2025 AGC–NCCER Workforce Survey, 92% of firms that were hiring reported difficulty filling open positions. Among salaried roles, estimators were one of the hardest to fill, named by 77% of firms looking for them. You can hire a new estimator. You can't hire 20 years of knowing which scopes always run long.
Some of that can be saved if it's written down while it's fresh. After each bid, a few notes on what the drawings missed, which quotes fell apart, and what the team would price differently give the next estimator something to build on.
That only happens when estimators aren't buried in takeoffs. Give them time back, and what they know stops walking out the door.
