Most firms still treat building code compliance as something that happens at the end of design. Plan check finds the issues. The team responds. The project moves on.
The cost of that pattern is hidden — until you measure it.
Two places to catch issues
On any given project, a code issue can be caught in one of two places:
- During design, while the model is still moving
- At plan check, after the set is sealed and submitted
Either path identifies the issue. They are not equivalent.
What plan check actually catches
Plan check is good at catching issues that survive the team's own review. It is not designed to catch them early.
By the time a set arrives at the AHJ:
- Layouts are fixed
- Consultant drawings are coordinated against them
- Specifications reference specific assemblies
- The project schedule is anchored to a submission date
A single comment on egress, occupancy, or accessibility at this stage rarely changes one drawing. It changes a layout, then a structural grid, then a stair, then the documents that reference them.
The hidden cost of late issues
Industry research on AEC rework is consistent: the cost of fixing an issue grows roughly an order of magnitude with each phase it survives.
A clearance question caught at SD is a comment. The same question caught in CDs is a coordination pass across multiple disciplines. Caught at plan check, it can hold a permit.
The cost is not only in hours. It is in:
- Carry costs on land or construction loans waiting for permit
- Schedule slip that cascades into other trades
- Fee erosion from unbudgeted revision cycles
- Trust with the client when a date slides
What 'during design' actually looks like
Moving compliance into the design phase does not mean reading more code earlier. It means letting the model itself surface the issues as it is being built.
In practice:
- Egress paths are evaluated against occupant loads as the plan changes
- Accessibility clearances are checked at every fixture, not just at the end
- Wall, floor, and roof assemblies are checked against the energy code as they are selected
- Each flagged issue is tied to the specific element that caused it
The team is still designing. The model is doing the cross-reference.
What changes for the team
The shift is less about new work and more about where existing work happens. Senior staff spend less time interpreting code on issues a junior could resolve with the right reference. Junior staff stop guessing.
Principals review fewer corrections sets. PMs hold fewer emergency standups two weeks before permit.
The hardest calls — the ones that require judgment about intent, precedent, and AHJ relationships — still need a person. The mechanical ones stop reaching that person.
What changes for the project
Submission becomes the place where compliance gets confirmed, not the place where it gets discovered.
The corrections set that used to define the weeks after submission shrinks. In several of the projects we have watched run this way, the first-pass corrections list moved from the typical dozens to the single digits — and in a few cases to zero.
The submission date and the construction-start date move closer together. That is the line that matters to the client.
Where this fits
None of this requires architects to stop being architects. It requires a system that can evaluate building code against the model continuously, in the background, without becoming another tool the team has to babysit.
We wrote about what that looks like here: AI building code compliance in Revit →
And the broader argument for why this has stayed manual for so long: Why building code compliance is still manual →
If you want to see how it changes a real project: Schedule a demo →