The building has to pass an inspection. The drawings that describe it never have to pass anything.
I have sat through plan sets from homebuilders, school districts, hospitals, and city halls, and I have never once seen two architecture firms lay a set out the same way. Not the sheet numbers, not what gets bookmarked, not how a room gets tagged, not even how a plain fact like a ceiling height gets written down. Everybody who has spent real time on the receiving end of these documents already knows this. What I never had a good answer for is why.
Construction runs on standards for almost everything that actually gets built, and the permit is what puts teeth in them. A government office checks the design against the code before anyone pours a foundation. Nothing in that process, or anywhere after it, checks how the drawing set itself is put together. A standard for that does exist, the US National CAD Standard, a consensus standard combining the American Institute of Architects' CAD Layer Guidelines, the Construction Specifications Institute's Uniform Drawing System, and the National Institute of Building Sciences' own plotting guidelines. I could not find anything that requires a firm to use it, or that checks whether they did.
There is exactly one place I could find where this actually gets enforced, and it explains the whole pattern. The federal government's General Services Administration requires the architects and engineers who prepare CAD deliverables for it to comply with the national standard, on top of its own additional CAD requirements, and states plainly that it will not waive either one under any circumstances. GSA can do this because it is the same buyer over and over, ordering enough buildings that it is worth the friction of making every architect who works for it draw things its way. I did not check whether every large repeat owner does something similar, only that GSA is the clearest example I could confirm of it actually being enforced. Almost nobody else hiring an architect is in that position. A hospital system doing one expansion, a city building one new hall, a homebuilder working with a firm for the first time, none of them have the repeat volume or the leverage to make demanding a format worth the fight. So the default just stays whatever the firm already does. Nobody is being lazy. Nobody with enough power over most individual jobs has ever had a reason to fix it.
And there is a real reason firms keep their own way of doing this, one that gets missed every time it comes up as a complaint. How a firm lays out its sheets, numbers its rooms, and organizes its own set is closer to a house style than an accident. It gets built over years by whoever is running the office, shaped by whatever software the firm came up on, and by the kind of buildings they actually specialize in. That is part of how a firm works, not incidental to it. This is exactly where I think a lot of construction AI gets it backwards. A tool that quietly expects every firm to reformat its own habits to match what the software wants is not asking someone to learn a new button, it is asking a firm to give up a piece of how it works just to be read by a machine. That is not stubbornness on the firm's part. That is a real cost, and it is a reasonable thing to push back on.
None of this would matter as much if the industry's actual fix for an unclear drawing were fast. In practice, it usually is not. The clearest measurement I could find of how long a formal request for information takes to come back, a 2013 analysis by the Navigant Construction Forum, put the median reply time at 9.7 days, rising toward 10 days on projects running two years or longer. That figure has a real limitation worth stating rather than hiding: it comes from one project-collaboration platform's own database, drawn mostly from projects in Australia and New Zealand, so it is evidence that the formal channel can run slow, not a verified number for how it runs on a typical American job. What I have not found a study for, and am arguing from my own experience instead, is how that same ambiguity plays out earlier, at the bid, where the deadline is usually tighter and does not move for anyone's answer. So every set that carries a real ambiguity quietly forces a choice nobody writes down anywhere: hold the bid and risk missing the window, or price it on your best read of what the architect meant and carry that risk yourself into the number you submit. That choice gets made on real jobs, all the time, by people who never say out loud that they are making it.
That's part of why I build the document-reading side of Setmark the way I do: teaching it to read a set's own conventions instead of assuming one firm's habits apply to the next.