The drawings knew.
Nobody had read them.

You can, and for a quick question across a few sheets it works well. We use those tools daily.

The difference shows up when the answer has to hold. A chat tool retrieves the parts of your drawings that look relevant to the question and answers from those, which is the right design for a conversation. It means the answer is drawn from a sample, and you cannot see how big that sample was. Resection reads for a different purpose: it accounts for every sheet, cites the exact text behind every fact, and stores the reading so the same question returns the same answer months later.

Curiosity and evidence are different jobs. If you are going to stake a notice on it, you need the second one.

It means the set is closed. Every sheet you upload is either read, or listed as not read with the reason, and that list is derived from the data rather than maintained by hand.

So when an answer says nothing was found, you can tell the difference between “nothing is there” and “nothing was in the part I looked at”. Every answer carries that boundary with it.

Follow the citation. Every fact carries the wording it came from, where that wording sits on the sheet, and the revision it was on, so you can put the drawing next to the answer and confirm it in seconds.

That is also the internal test: anything the system cannot tie back to specific source text is rejected rather than stored. You should never have to take an answer on trust, because a claim you cannot check is not worth having in a commercial argument.

Yes. The reading is done once and kept, and later questions run against that stored record rather than re-reading from scratch.

This matters most at the point it is challenged. An assessment you made in the spring can be produced again in the autumn, with the same citations, rather than being re-derived and coming out slightly differently.

No, and it never will. Resection checks consistency and coordination: whether two documents say the same thing, whether a reference points at something that exists, whether a revision changed something that affects you.

Whether the design is adequate is engineering judgement, and that stays with your engineer. We are careful about this line because a tool that blurred it would be dangerous.

It says so, in the answer, rather than filling the gap with something plausible.

Scanned sheets, hand mark-ups and drawn detail with no written label are all real. Where something has not been read it appears on the list of what was not covered, instead of quietly shrinking the set the answer came from.