Deloitte’s 2026 Engineering & Construction Industry Outlook puts the cost of persistent labor shortages at nearly $124 billion in lost output, with the industry needing roughly 500,000 additional workers this year. The number gets quoted as a field problem. A bidding-software founder writing in ENR argues the loss starts earlier than that, in preconstruction.
Where construction output actually gets decided
Shiva Dhawan’s argument is simple enough to check against your own last quarter. Output doesn’t begin when a crew mobilizes. It begins when somebody turns drawings into quantities, quantities into costs, and costs into a bid. When estimating runs out of capacity, projects don’t slip. They never get pursued, never get won, and never become construction volume at all.
Every contractor has a version of this: an invitation to bid that aged out in an estimator’s inbox, or a qualified opportunity passed on during a busy month because nobody had the hours to price it. That work is invisible in every metric the firm tracks. It doesn’t appear in backlog, win rate or lost-bid analysis, because it was never a bid.
Why the labor-shortage framing misses precon
The Deloitte figure assumes output is constrained only when craft labor isn’t available. That’s a reasonable model for a market where demand is limited by hands on site. It’s a poor model for one where a contractor with a healthy backlog declines qualified work because precon is the bottleneck.
Estimating has rarely been treated as a strategic capacity function. Contractors track billable field hours, equipment utilization and crew productivity closely, and track estimator throughput barely at all. The staffing level for precon tends to be whatever it was last year.
What to measure before buying software
Dhawan sells AI bidding tools, so read the prescription accordingly. The underlying diagnostic doesn’t depend on the product. Count the ITBs your firm received last quarter and the number you actually priced. Track the ones declined for capacity rather than for fit, geography or risk. If that second number is material, you have a precon capacity problem, and hiring another superintendent won’t touch it.
Whether the answer is automated takeoff, more estimators, or better filtering on which invitations deserve attention is a separate question, and the answer differs by trade. Self-performing subs with repetitive scopes get more out of automation than a GC bidding one-off institutional work. But the measurement costs nothing, and most firms have never run it.