The crew poured the wrong footing. Not because anyone was careless — because the foreman was building from a set he printed two weeks earlier, and the foundation plan had since gone to Rev C. The structural engineer had shifted a column line eighteen inches to clear a utility nobody knew was there until the locate came back. The change went out by addendum, got logged in the office, and never made the trip to the gang box by the slab. So a four-man crew did everything right against a drawing that was already wrong. That gap — between the revision that exists and the revision the crew is holding — is the entire reason document control exists.

Every project runs on documents that change. Plans get revised. Specs get clarified. Contracts get amended by change order. RFIs get answered. Submittals get stamped, rejected, resubmitted. None of that is failure — it is how a building actually gets designed, because nobody knows everything on day one. The job of document control is to make sure that out of all the versions of a thing that have ever existed, the team builds from the one that is current and approved. It is also the source of a remarkable amount of expensive trouble. If you want the one-line definition, the glossary has it — this is the longer story.

What document control actually is

Document control is the practice of keeping every project document current, approved, distributed, and traceable — so the version a person acts on is the version they were supposed to act on. Strip away the software and it is three promises. First, there is one identifiable current revision of each document, and anyone can tell which one it is. Second, when a revision supersedes the old one, the old one stops being used — pulled, archived, or clearly marked dead. Third, you can reconstruct who had what, when, and why a change was made, long after the fact. That third promise — traceability — is the one people forget about until a dispute lands eighteen months in, and “the column moved” becomes “moved when, approved by whom, carried by which addendum, and was the crew building off the right sheet when they poured?”

Why it exists: the field builds the current revision

A drawing set is a snapshot of intent at one moment, and the moment it is issued it starts aging. The contract documents anticipate this — that is why drawings carry a revision block, why addenda are numbered, why a change order is a formal instrument and not a text message. The whole apparatus exists so a hundred people across a dozen trades can be pointed at the same moving target and stay aligned as it moves. Here is the load-bearing idea: a revision is not just a newer file, it is an instruction that the previous file is no longer authoritative. Being right in the office is worthless if the field is framing off Rev B.

Rev C doesn’t sit politely next to Rev B. It kills Rev B. A superseded sheet still in someone’s hands isn’t an old version — it’s a wrong instruction with a real address.
On why supersession matters
One change rarely stays in one document — control is what keeps the connected pieces from drifting out of sync.

The three classic failure modes

Version drift is the slow one. The architect issues Rev C as a partial set — six sheets, not the whole package — and emails it to the GC’s project manager. The PM forwards it to the super, who thumbs through it on a phone in the field. The drywall sub never gets it, because they weren’t on that thread. Now four people hold four slightly different versions of the truth and nobody knows it, until the framing inspector points at a wall that isn’t where the current plan puts it. Superseded sheets are the sharp one: a revised detail goes out, but the old printed set is still pinned to the trailer wall and tucked in the gang box. The new detail changed a flashing condition; the crew, reasonably, builds what’s in front of them. The error stays invisible until water shows up in that wall two winters later — a callback, a finger-pointing meeting, and a bill for wet insulation. Then there’s the orphaned answer: an RFI gets a clear, correct response buried in an email reply, and it never gets written back onto the drawings or logged against the spec section — so six weeks later a different sub asks the same question and gets a different answer. In all three, the right information existed the whole time. The breakdown was in distribution and connection, not storage.

From filing documents to reading them

Traditional document control treats a drawing as an opaque object — a file with a name, a revision tag, and a distribution list. The system knows the sheet is A-301 Rev C. It does not know that Rev C moved a column, or that the move now contradicts a dimension still printed on S-201, or that RFI #22 was the reason for the change in the first place. A human has to open the file, read it, hold the rest of the set in their head, and catch the conflict — and on a busy job, sometimes nobody does. The discipline changes when the system can read the documents instead of just shelving them. Once the contents are understood, the set stops being a stack of files and becomes a connected record, where a revision ties back to the question that prompted it and the section it touches.

The shift isn’t a better folder — it’s the difference between knowing a file exists and knowing what it says.

This is the lane BRAD works in. You forward it the things that already run the job — the plans, specs, contracts, change orders, RFIs, submittals, the field photos, the email and text between office and field — and it reads them and connects them into one record. Then the team asks questions the way they already talk, over email or text, and BRAD answers from the current set with the citation attached: not “the tile is Rev C,” but “the floor tile spec is 09 30 13, revised by change order #14 — here’s the sheet.” The answer lands where the crew already looks, with its receipt, so nobody has to take it on faith.

Document control has always been a promise the field makes to the drawings: we will build what’s current and approved, not what we happened to print. The hard part was never wanting to keep that promise — it was knowing, on any given day, which sheet was actually current and where the answer to a given question last lived. Read the documents instead of just filing them, and that knowledge stops being something a person has to carry in their head. The set still belongs to the people who stamp and sign it. They just stop having to dig for it.