I set a specific quality bar for the virtual architecture bench we were building, structural, mechanical, civil, plumbing, HVAC, fire, landscape, jurisdictional review, all feeding into the plans our system produced. The bar wasn't internal satisfaction. It was that the third party stamping engineers who actually review our plan sets should come back with zero red lines, and I deliberately chose not to let the system self stamp anything, keeping a licensed human as the seal on every set regardless of how good the tools got. My reasoning was that we were building this capability to run projects forever, not to win one job faster.
My opinion is you measure quality by the reviewer's own red pen, not by how confident your own team feels walking into the review. A team that grades its own work will always feel more confident than the results justify. The stamping engineer's red line count is the only scoreboard that can't be talked into a better mood.