The problem
Every project meant collecting vendor bids in inconsistent formats and manually leveling them: lining up scope, catching what each bid quietly excluded, and chasing vendors with follow-up questions. That took the team two, three, sometimes four weeks per cycle — senior people doing spreadsheet archaeology while the schedule waited on the answer.
What we built
Delivered with Last Rev: an AI bid-leveling workflow that ingests the incoming bids and produces a detailed leveling analysis — vendor-by-vendor completeness against the bid scope, flagged gaps and exclusions, and automated follow-up questions for each vendor where the bid was silent or ambiguous. What was a manual process became a reviewed, repeatable one.
The bid library
We also built a library of the firm's historical bids, so every new bid lands in context. The team can now see which vendors have been more accurate on cost estimates over time, and which ones habitually underbid and change-order their way back. Vendor selection moved from memory and relationships to historical data.
The result
Bid leveling that took two to four weeks now takes minutes, with the detail level of a senior estimator's workup: completeness analysis, follow-up questions drafted, and historical vendor accuracy alongside every number. The estimators still make the call — they just make it with the assembly work already done.