In Surrey, British Columbia, city officials ran the numbers on their residential permit applications. 80% contained significant zoning deficiencies. Not minor formatting issues. Significant violations — setback encroachments, floor area ratio overages, height limit breaches. Each deficient application required an average of 1.6 resubmissions before approval.
Resubmission is a euphemism. It means your designer fixes the plans, your builder waits, your lender charges interest on money that isn’t building anything, and a reviewer who already read your plans once reads them again. Multiply that by every application in the queue.
Surrey decided to stop complaining about the problem and handed it to a machine.
How Slow Is Slow
The national average building permit initial review time in the United States is 22.9 days (mean) and 14 days (median), according to guidelines published by 741 U.S. cities. Those numbers represent the first review only. After corrections, resubmissions, and interdepartmental routing, the real timeline for a residential permit typically runs two to four months. In some jurisdictions, much longer.
Honolulu’s median wait for a commercial permit is 393 days. More than a year. The city is still running permitting software from the 1990s.
NAHB’s most recent analysis puts the regulatory burden at 23.8% of the average new home sale price — $93,870 on a $394,300 home. That includes $7,640 in building permit fees, $6,260 in water and sewer inspection fees, $6,480 in architecture and engineering costs, and $6,367 in impact fees. Nearly all developers surveyed (95.9%) reported that regulatory compliance alone added approximately six months to their project timelines.
Six months. On a construction loan charging 8–12% interest, that’s $15,000–$24,000 in carrying costs on a $300,000 draw — money that buys nothing and builds nothing.
Where the Reviewers Went
Building departments aren’t slow because they enjoy it. They’re slow because they’re understaffed and getting worse.
In Utah, 60% of building inspectors are close to retirement. Eighty-five percent are over 45. Only 3% are under 35. George Williams, a Salt Lake City inspector, described being “without fail” the youngest person at every ICC training session — “not slightly younger, but dramatically younger than nearly everyone else.”
The pipeline isn’t refilling. Government inspection jobs lost their competitive edge years ago. Salaries lag the trades. Benefits eroded. Job security isn’t what it was. A competent plans examiner can make more money as a contractor, with fewer angry applicants in the lobby.
So departments run short-staffed, review queues grow, and the applicants who submitted clean plans wait behind the applicants who didn’t — because nobody told them their plans were deficient until a reviewer spent two hours finding out.
Spell-Check for Zoning
Surrey’s solution was Archistar’s AI PreCheck, which went live for public use on March 10, 2026. The city’s business transformation manager, Jerome Thibaudeau, described it as “spell-check for zoning.”
Designers and homeowners upload building plans. The system scans them against local zoning rules — setbacks, height limits, lot coverage, floor area ratios — and returns a compliance report flagging violations. Fix the issues before you submit. Arrive at the counter with a clean application.
It’s not making the decision. The plan reviewer still approves or rejects. But if 80% of applications previously arrived with significant deficiencies, and most of those deficiencies are checkable against known numeric rules, then catching them before submission eliminates most of the resubmission cycle.
Ron Gill, Surrey’s general manager of planning and development, called it “a major step in modernizing how we support home designers and applicants.” The initial rollout covers single-family homes in R3 zones, expanding to duplexes, coach houses, garden suites, and houseplexes.
Austin, Los Angeles, Honolulu
Surrey isn’t alone. Austin formally adopted Archistar in October 2024 after a three-month pilot, starting with residential plan review. Stephanie Sanchez, a senior public information specialist for the city, noted that “there haven’t been critical errors or unexpected behaviors” from the platform during testing.
Los Angeles County and the City of Los Angeles launched the same system on April 30, 2025, driven by wildfire recovery. Governor Newsom signed an executive order on January 12 to speed permitting for fire-damaged properties, explicitly citing the need for automation. “The current pace of issuing permits locally is not meeting the magnitude of the challenge we face,” he said.
Honolulu, staring down its 393-day median and its 1990s-era software, announced AI adoption in June 2025. The city partnered with Vancouver-based Clariti for workflow management and Chicago-based CivCheck for AI code compliance scanning.
CivicPlus, a public-sector SaaS company serving over 13,000 government agencies, announced a partnership with CodeComply.AI on March 19, 2026. CodeComply.AI automates compliance checks against IBC, IMC, IPC, IRC, ADA, FHA, UFC, and multiple NFPA codes. The fine print: automated checks are “informational and do not replace professional review or constitute legal advice.”
What AI Actually Checks
These tools are pattern matchers, not structural engineers. They excel at dimensional compliance: is the proposed structure within the setback? Does the lot coverage exceed the zoning limit? Is the building height under the cap? Does the stairway width meet IRC minimums?
They handle code lookups fast. A human reviewer checking a residential plan against the IRC has to hold dozens of provisions in their head — egress window sizing (R310.2), smoke alarm placement (R314), stairway geometry (R311.7), guard height (R312.1), energy compliance (N1101–N1108). An AI system checks all of them simultaneously and consistently.
Where they struggle: judgment calls. An unusual lot shape that technically violates setback requirements but has a variance in process. A structural configuration that meets code but raises engineering concerns. A historic district overlay with subjective design review standards. These still need a human who understands that code enforcement involves interpretation, not just measurement.
| What AI Plan Review Does Well | What It Doesn’t Do |
|---|---|
| Dimensional compliance (setbacks, heights, FAR) | Structural engineering review |
| Code section lookups and cross-references | Geotechnical assessment |
| Egress, fire separation, accessibility checks | Variance and exception processing |
| Pre-submission deficiency screening | Subjective design review (historic, aesthetic) |
| Version comparison between resubmissions | On-site physical inspections |
CivCheck’s 80% Claim
CivCheck, acquired by Clariti in October 2025, markets itself as capable of reducing “overall permit approval times by 80%+.” That number deserves dissection.
If a permit currently takes 90 days and AI review cuts it to 18, that’s transformational. But the 90 days typically break down into: application intake (1–3 days), plan review routing (3–5 days), plan review itself (10–20 days), corrections and resubmission (20–40 days), interdepartmental reviews — fire, health, environmental, public works (10–30 days), and final approval processing (3–10 days).
AI plan review addresses maybe 30–40% of that timeline: the initial review and the corrections cycle. Cutting those by 80% is plausible. Cutting the entire permitting timeline by 80% requires automating parts of the process that these tools don’t touch — fire marshal review, environmental clearance, interdepartmental sign-offs, and the simple organizational dysfunction of an underfunded municipal office.
That said, corrections and resubmission are often the largest single block. If Surrey’s data is representative — 80% of applications arriving deficient, averaging 1.6 resubmissions — then eliminating most of that cycle alone could cut 20–40 days off a typical residential permit.
What a Homeowner Can Do Right Now
If your building department doesn’t offer AI pre-check (and most don’t), the practical lesson is simpler than the technology.
Hire a permit expediter. In major metros, these range from $2,000–$5,000 for residential. They know the local code, the local reviewers, and the local pitfalls. Their entire job is submitting clean applications that don’t bounce.
Or, if you’re working with an architect or designer, ask them explicitly: “Have you run this against the current zoning ordinance?” Not the building code — most competent architects handle that. Zoning. Setbacks. FAR. Lot coverage. The stuff that varies by municipality and changes every few years. The stuff that accounts for 80% of deficiencies in Surrey.
If you’re a production builder submitting 50+ permit applications a year, contact CodeComply.AI, Archistar, or Clariti directly. Even if your jurisdiction hasn’t adopted AI review officially, running your plans through a pre-check before submission could eliminate most of your resubmission cycles.
Strongest Counterargument
Plan review isn’t always the bottleneck. In many jurisdictions, the real delays are inspections, environmental review, utility coordination, and the political process of getting a planning commission to approve your project. A tool that cuts plan check from 14 days to 2 days doesn’t help much if the total timeline is 180 days and plan check was only 8% of it. The 83% of developers who cite “permitting delays” as a barrier to housing development aren’t distinguishing between plan review delays and everything-else delays. AI plan review is a real improvement to one part of a multi-part problem. It is not a permit in a box.
Limitations
No published data quantifies actual time savings from Archistar deployments in Austin or Los Angeles — both adoptions are too recent for meaningful outcome studies. CivCheck’s 80% reduction claim is a marketing figure from a privately held company with no independent verification. Surrey’s 80% deficiency rate may not be representative of U.S. jurisdictions, which have different code structures and application standards. NAHB’s regulatory cost analysis uses 2021 data; the regulatory burden may have shifted since. Honolulu’s 393-day figure is for commercial permits, not residential. The building inspector shortage data comes primarily from a 2017 CityLab report and Utah-specific statistics; national comprehensive data on inspector demographics is limited. CodeComply.AI supports multiple code families but does not publish accuracy rates or false-negative data for its automated compliance checks.