Every preconstruction team has the same unwritten policy: we bid everything until Thursday afternoon, and then we discover we have bid nothing well. The invitation inbox never got the memo that the estimating team is three people.

That tension just got a number attached to it, and it is a useful starting point for the question of which bids deserve your scarce estimator hours.

When estimating capacity is the bottleneck, decide bid or no-bid by expected margin per estimator hour, not by how interesting the project looks. Score every invitation on fit with your trades, your history with the owner, the likely field of competitors, and the hours the bid will realistically consume, then price only the top of that list. AI belongs in the second half of that process, shrinking the hours each chosen bid costs, rather than replacing the judgment about which bids to chase.

Why is estimator capacity the real constraint?

A sponsored piece on Construction Dive, supplied by the vendor Beam AI, frames estimating capacity as the core bottleneck in preconstruction. It pairs growing project pipelines with a Bureau of Labor Statistics projection that cost estimator employment will fall about 4% from 2024 to 2034, even as construction employment overall is projected to grow. Because it is sponsored content, read the argument as vendor positioning, but the BLS projection is the part worth keeping. Other coverage reads the same figure as a profession in a generational handoff, with openings driven largely by retirements and replacement demand.

The practical reading for a GC: demand for pricing is rising while the pool of people who can do it is not. Something has to give, and the cheapest thing to give up is the bids you were never going to win.

How do I build a bid or no-bid score that actually works?

Keep it to four or five factors you can fill in within minutes: does the scope match trades you self-perform or know well, do you have a relationship or past win with this owner or architect, how many bidders are likely, is the schedule realistic, and how many estimator hours will it take. Then track the outcome. The score is only as good as your own history, which is why it pays to log hours per bid and hit rate by client type before you buy anything.

This is where teams skip a step. They buy a takeoff tool to bid more, when the better first move is to bid fewer of the wrong ones. AI that does the measuring faster makes a good bid cheaper. It does nothing for a bad bid except make it cheaper to lose.

Where does AI takeoff genuinely buy back estimator hours?

Quotr, a Berkeley startup, announced on September 30, 2026 a $4 million seed round led by Llama Ventures. According to the coverage I found, its platform reads uploaded PDF or image plan sets, measures lengths and areas, counts quantities across sheets, and then produces a priced estimate and proposal. The company says contractors using it have cut takeoff time by up to 80%, roughly from 20 hours to one or two. That is a company-reported figure, and I could not verify it independently. Reported use of funds includes enterprise rollouts, growing its supplier network, and expanding into heavy industrial and data center work. Coverage also disagreed on some co-investor details, so treat the finer points of the round with caution.

If even half of such a claim holds on your drawings, the math changes the no-bid threshold: a bid that cost 20 hours and now costs 8 can clear your margin-per-hour bar where it did not before. That is the honest way to use these tools in the bid decision. Re-run your own score with the new hour estimates, rather than assuming you can now bid everything.

What should still stay human?

One analysis from DCD argues that AI does not improve the accuracy of cost baselines, which still depend on historical project data. I only saw that in a search summary, but it matches what estimators say in practice: the machine can count the doors, and the person decides what the doors cost in this market, with this sub, on this site. Keep the review step, keep the audit trail, and keep the decision to bid or walk with a human who owns the number.

The firms that handle a shrinking estimator pool well will probably not be the ones that bid the most. They will be the ones that know, in ten minutes, which five invitations out of twenty deserve the team's week.

Frequently asked questions

How do I decide which bids to skip when I do not have enough estimators?

Score each invitation on fit, client history, competition, and the estimator hours it will consume, then bid only the ones where expected margin per estimator hour is highest. Use AI to cut the hours a bid costs, not to decide whether it is worth bidding.

Will AI takeoff tools solve the estimator shortage?

They can shrink the repetitive measuring work, and vendors such as Quotr claim takeoff time cuts of up to 80 percent, but that is a company claim. The estimator still owns judgment, scope review, and the final number, so treat AI as added capacity for the first pass.

What data do I need to make a good bid or no-bid call?

At minimum, your hours spent per bid, your hit rate by client and project type, and your realized margin on past wins. Without that history, any AI scoring is guessing.

Sources