AI for general contractors that reads every document on the job

A general contractor does not build the building. The GC coordinates the people who do, which means the job is run on documents and the small questions about them that arrive all day. A couple dozen subs, a few thousand pages of plans and specs, and a steady drip of “which revision is current” and “was that change ever approved.” Most of those questions already have an answer somewhere in the record. The cost is the time spent finding it. US construction teams lose about 5.5 hours a week just looking for project data, before anyone fixes a thing.^[1] Brad reads everything your jobs generate and gives the answer back in seconds, with the document behind it.

This is not a niche tax. Construction productivity has barely moved in decades while the rest of the economy pulled ahead, and a large part of the gap is information that should be at hand but is not.^[2] The plans say one thing, the spec says another, the answer is in an RFI from March that lives in someone’s sent folder. A GC carries that coordination load on every job at once. Brad is built to carry the part of it that is pure retrieval, so your PMs and supers spend less of the day playing detective and more of it running the work.

Built for the coordination load

As the GC, you are ball-in-court on every document on the job: plans and specs, subcontracts, change orders, the submittal log, invoices, and the RFIs and field texts that ricochet between the trailer and the wall. Brad holds that whole record and, more to the point, connects it. A subcontract ties to the scope it covers, a submittal to the spec section it answers, an RFI to the detail it questions and the change order it triggers. Construction ranks near the very bottom of every industry on digitization, which is a polite way of saying most of this record still lives in PDFs, email, and people’s heads.^[3] Brad reads it where it is.

Ask where a project stands and the answer is one question away. “What’s still open on the curtain wall?” “Which submittals are holding the long-lead steel?” You get the list back with a source on each line, instead of a half-hour spent dredging your sent folder like an archaeologist. The retrieval that used to be a person’s afternoon becomes a sentence.

Answers for the office and the field

The office and the field ask different questions. Your PM needs to know whether a change order burned float on the critical path. Your super needs the current spec at the wall before the crew sets it against a revision that is two weeks stale. Brad answers both, over the email and text your team already uses, and every answer comes back with the document behind it. There is no new app for the field to learn and no portal to log into from a truck.

That last part matters more than it sounds. Waiting on information is non-value-adding work in the lean sense: the crew is on the clock, the activity cannot close, and nothing is being built while the question sits.^[4] A callback that lands like a voicemail nobody returns is float quietly leaking off the schedule. When the answer arrives in seconds with the spec section attached, the field keeps moving instead of standing around, and the super is checking a sourced answer rather than guessing from memory.

Across every job, not just one

GCs run several jobs at once, each with its own cast and its own fires. Brad keeps each one as its own connected record, sealed off from the others. You can ask about any active job and get an answer grounded in that project’s documents and nobody else’s. One site’s spec section never wanders into another’s answer, and a sub on three of your jobs is not a leak between them.

This is the difference between a project brain and a search box. A search box hands you ten PDFs and wishes you luck. A project brain knows that CO-014 on the Maple Street job amended the schedule of values, which changed the retainage math on the next pay app, and it keeps that chain inside the right project. Across a portfolio, the same plain-language question works on every job without you re-learning where each one’s files are buried.

It connects the money to the work

Change orders, the schedule of values, retainage, and the invoices against them are where margin is won or quietly bled away. Bad data is not a rounding error here: poor information cost global construction an estimated 1.85 trillion dollars in 2020, and roughly 88.7 billion of that was rework traced directly to it.^[5] A charge that does not trace to an authorized change is exactly that kind of leak.

Brad links the money to the work. A charge traces back to the change order that authorized it, CO-014 to the invoice line that billed it, and a scope change surfaces the cost and the float it burned while you can still do something about it. That timing is the whole point. The ability to influence cost falls steadily as a project moves forward, so a cost question answered in week three is worth far more than the same question answered at closeout.^[6] A surprise that surfaces at closeout is a surprise you pay for.

The line between Brad and your judgment

Brad is document intelligence for construction, not a guarantee or a certification. It reads, connects, and drafts from your project’s documents and messages, and it answers with the source attached. What it does not do: replace a PM’s judgment, sign off on a change order, or stand in for the licensed reviewer who owns a design answer. A person approves the decision that goes back to the field every time. When the documents genuinely disagree, Brad flags the conflict and points at both sources rather than papering over it with a confident guess, because a discrepancy between the drawings and the spec is a human’s call.

Your project’s content is yours and stays sealed inside your workspace, walled off from every other one. If you have specific requirements about how project data gets handled or retained, ask us and we will walk you through exactly how it works, in plain terms.

You cannot stop the questions from arriving, and you would not want to. The questions are the job. What you can change is how long the answer takes and whether it comes back with its source or as one more guess waiting on someone’s memory. Brad reads every document on the job so your team stops being the search engine and gets back to running the work.