A general contractor's estimators were spending 15 hours leveling each set of subcontractor bids by hand. Now it takes about 20 minutes, a 98% reduction, and the estimator still makes every call.
The problem
Sub bids arrive in every format (PDF, Excel, Word) and every structure (turnkey, itemized, or a single lump sum). Before any judgment could happen, the team spent about 15 hours per package pulling numbers out of each proposal, lining up scope, and building the side-by-side by hand. Slow, and easy for a buried exclusion or a mismatched plan date to slip through.
What we built
A tool the estimating team runs themselves. It reads every bid in any format, lays them out trade by trade, and shows, item by item, whether each scope line is included, excluded, priced separately, or not addressed. Then it flags what deserves attention:
- Scope gaps, where subs disagree on what is in or out, or an item nobody priced.
- Commercial risks, short bid validity, aggressive change-order language, tariff pass-through clauses.
- Questions for the estimator, the judgment calls only a person can make, like "does Sub A's turnkey price already include the low-voltage scope Sub B priced separately?"
It never picks a winner and never touches the budget. AI gathers and flags; the estimator reviews and decides.
The result
Fifteen hours down to about twenty minutes per package. The estimator opens a finished comparison with the exceptions and questions already surfaced, and spends the time on the judgment that decides the buy instead of assembling spreadsheets.
