Members are the easy part
Getting a frame modelled — columns, beams, bracing, the grid — is largely mechanical once the design is settled. It is repetitive, it is well understood, and it is the part most teams already move through quickly.
That is also why it is the wrong place to look for savings. If the frame takes a fifth of the time, halving it changes very little about the programme.
Connections are where the hours accumulate
Every joint is a decision, and the decisions interact. What sits where, what clears what, which of your shop's preferred details applies here and why the one next to it is different. Multiply that across a frame and the arithmetic stops being about modelling speed at all.
It compounds when the same conditions are detailed slightly differently by different people, because then review has to catch the inconsistency rather than just checking the work.
Rework is the part nobody budgets for
The hours that hurt most are the ones spent twice. A condition that does not fabricate cleanly, an erection sequence that will not go together in that order, a clash found after drawings were issued. None of that appears on an estimate, and all of it lands on the same people already on the critical path.
This is the argument for checking early rather than checking thoroughly. A problem found in the model is a change; the same problem found on the shop floor is a cost.
What to automate first
The order that tends to hold: automate the connection work against your own standards, then check the result before it is released, then price from the same model rather than a re-measured copy.
The reason that order works is that each stage makes the next one cheaper. Automating connections without checking them moves the problem downstream; checking a model nobody trusts is not much use either.