Subcontract review from 2–3 hours to 15 minutes for a commercial masonry contractor.
A regional commercial masonry contractor was losing 2–3 hours — or several hundred dollars in legal fees — reviewing every long-form subcontract before signing. We built an AI tool that reads the full agreement (up to ~300 pages), flags the risky clauses, and produces a lawyer-ready summary in about 15 minutes.
In commercial construction the sub never writes the contract — the GC sends a take-it-or-leave-it agreement, often 100–300 pages with exhibits and flow-down clauses, on a tight deadline. Reviewing one properly meant 2–3 hours of a principal's time or several hundred dollars to an attorney, so contracts often got skimmed and signed unread.
An AI review tool that ingests the full PDF (exhibits included), analyzes the clauses that carry risk for a trade sub — payment timing, retainage, indemnification, insurance, termination, flow-down — and returns a plain-English summary plus a short, lawyer-ready list of the clauses worth a second look. It targets legal judgment instead of replacing it.
Contract review dropped from 2–3 hours to about 15 minutes. Attorney time is now scoped to the flagged issues rather than the whole document, bids turn around faster, and every contract gets read in full the same way.
TL;DR
A regional commercial masonry contractor was signing subcontracts they didn’t have time to fully read. Each general-contractor agreement ran anywhere from a few dozen to nearly 300 pages of terms, exhibits, and flow-down clauses — and every one carried real money in its payment terms, retainage, indemnification, and insurance requirements. Reviewing one properly meant 2–3 hours of a principal’s time, or several hundred dollars to send it to an attorney, and often both. We built an AI tool that reads the entire agreement, flags the clauses that actually matter, and produces a plain-English summary plus a focused issue list ready to hand to a lawyer — turning a 2–3 hour bottleneck into about 15 minutes.
The problem
In commercial construction, the subcontractor almost never writes the contract. The general contractor sends a take-it-or-leave-it agreement, and the sub has a narrow window to review it, flag anything unacceptable, and sign — or lose the job. These are not short documents. A single subcontract routinely arrives with the base agreement, a stack of exhibits, insurance and bonding requirements, and “flow-down” clauses that silently bind the sub to terms buried in the prime contract they’ve never seen.
The client’s review process didn’t scale with their pipeline:
- A principal read each contract by hand — 2–3 hours of the most expensive person in the company, doing it after the workday because there was no time during it.
- Or it went to an attorney — several hundred dollars per review, and a multi-day turnaround that could jeopardize a bid deadline.
- Usually both, which meant the risky path was skimming: signing agreements without anyone having truly read the indemnification language, the payment timing, the retainage terms, or what the flow-down clauses actually pulled in.
The real cost wasn’t just the hours or the legal invoices. It was the exposure — agreements getting signed because reading them properly was too slow, and the occasional job passed on entirely because the review couldn’t happen before the deadline.
What we built
An AI contract-review tool built around one job: read the whole thing, faster and more consistently than a tired human at 9pm, and surface what a decision-maker needs to see.
- Full-document ingestion. Upload the subcontract PDF — including the long exhibits most tools choke on — and the system extracts the complete text, not just the first few pages.
- Clause-level analysis. The document is analyzed against the terms that carry risk for a trade sub: payment and pay-when-paid timing, retainage, indemnification and “hold harmless” language, termination and default triggers, insurance and additional-insured requirements, liquidated damages, change-order procedures, and flow-down provisions.
- A plain-English summary. Instead of 300 pages, the principal gets a readable brief: what this contract says in normal language, and where it departs from what’s standard or reasonable.
- A lawyer-ready issue list. The output isn’t “trust the AI.” It’s a short, specific list of the clauses worth a second look — so when an attorney is worth involving, they’re reviewing the three things that matter, not re-reading the whole document at their hourly rate.
The design principle throughout: the tool doesn’t replace legal judgment, it targets it. A human still makes the call — they just make it in 15 minutes with the risky clauses already in front of them, instead of 3 hours in, hoping they didn’t miss something on page 214.
The results
- 2–3 hours → about 15 minutes per contract review.
- Legal spend scoped to the flagged issues rather than a full-document read — the attorney reviews the short list, not the whole agreement.
- Faster bid and signing turnaround, because review stopped being the step that couldn’t happen in time.
- Consistent coverage — every contract gets read in full, the same way, instead of depending on how much time was left at the end of the day.
Why it matters
This is the shape of work AI is genuinely good at right now, and it’s a useful counter to the hype: it isn’t a chatbot bolted onto a website. It’s a specific, judgment-heavy, document-heavy task — the kind of thing a small business either does slowly and expensively or quietly skips — handed to a model that reads carefully and reports back, with a human still holding the final decision.
That’s the pattern we look for in an AI engagement: a real bottleneck with a clear before-and-after, where the model does the reading and the person keeps the judgment. This one worked well enough that it shaped how we think about contract-review tooling for the trades more broadly.