A Virginia homebuilder submitted plans for a single-family residence to its county building department in early 2026. Weeks passed, then a month, then two. When someone finally called to ask about the status, the answer was brief and devastating: the staff member handling the review had resigned, the application had not been reassigned, and it was sitting in a digital queue attached to a login nobody was monitoring.
Nobody lost the plans, and nobody rejected them. They simply stopped moving through the system because the one person responsible for moving them was gone, and the institutional machinery had no fallback.
This is not unusual; it is, in fact, the default condition of residential permitting across most of the United States.
A Workforce That Is Aging Out
The International Code Council surveyed its membership between 2014 and 2017 and found that fifty-four percent of code officials were over the age of 55. That survey is now nearly a decade old, the ICC has not published an update, and many of those officials have since retired.
SAFEbuilt, which provides outsourced building department services to municipalities, calls it the “silver tsunami”: losing one or two experienced reviewers can mean immediate backlogs and friction with developers. New hires take years to reach full competency, because code interpretation requires judgment built through thousands of applications across dozens of edge cases, and that judgment walks out the door with every retirement.
What the Delay Costs You
Permitting delays are not free. Every month your project sits in a review queue, you pay to hold the land, service the construction loan, and cover insurance on a property generating no revenue.
The City of Austin quantified this in a 2023 site plan review report: a single-family home redevelopment carries holding costs of $9,700 per month during the permitting phase, and a multi-family residential project valued at $38 million costs $546,000 per month in carrying charges alone.
Washington state’s Building Industry Association conducted a statewide study and found the average permit delay is 6.5 months, adding $31,375 in holding costs to each home. Using NAHB estimates that every $1,000 added to construction costs prices out 2,200 families from homeownership, that single delay pushes 69,025 Washington families past the affordability threshold.
The Fix That Already Works
Honolulu’s Department of Planning and Permitting was, for years, a cautionary tale: building permits sat in limbo for months, Civil Beat documented the dysfunction in exhaustive detail, applicants complained publicly, and developers threatened to take projects elsewhere.
Then the department did two things that changed the trajectory entirely. In 2022, robotic process automation pre-screened applications for completeness, bouncing incomplete submissions instead of letting them clog the pipeline; wait times dropped from six months to days. In 2023, a partnership with CivCheck brought AI into the code compliance review itself. CivCheck does not approve or reject permits; it reads plans, flags potential compliance issues, highlights relevant code sections, and presents reviewers with a structured summary instead of a raw PDF.
Review time per application dropped from 60 to 90 minutes to 15 to 20 minutes. That is not a typo. The same work, on the same applications, completed in a quarter of the time, because the AI handled the tedious cross-referencing and the human handled the judgment.
“Our AI doesn’t make decisions,” CivCheck CEO Dheekshita Kumar told GovTech. “It helps humans make better decisions faster by organizing information, flagging relevant code sections, and surfacing issues that still require professional judgment.”
Archistar, an Australian company, takes a complementary approach with its AI PreCheck platform. Computer vision and machine learning parse architectural drawings, mapping setbacks, heights, lot coverage, and energy specifications against a jurisdiction’s exact zoning and building regulations, producing a standardized compliance report in minutes. Austin has formally adopted it as a pre-check layer for single-family permits. Govstream.ai serves Bellevue, Washington. Louisville, Kentucky and Naples, Florida are also deploying AI permit tools. These are production systems processing real applications, not research prototypes.
What the National Math Looks Like
The U.S. Census Bureau recorded approximately 1.5 million residential building permits issued in 2024. If 30 percent experience delays averaging two months at $5,000 per month in holding costs (conservative, given Austin’s $9,700), that is $4.5 billion in annual carrying costs passed to homebuyers. Cutting delays in half saves $2.25 billion per year.
| City | AI Platform | Pre-AI Wait | Post-AI Wait | Status |
|---|---|---|---|---|
| Honolulu, HI | CivCheck | 6 months | 7 days | Production |
| Austin, TX | Archistar PreCheck | Months | Minutes (pre-check) | Production |
| Bellevue, WA | Govstream.ai | Weeks | Reduced | Production |
| Louisville, KY | TBD | N/A | N/A | Pilot |
| Naples, FL | AI plan review | Months | Reduced | Production |
Why Most Cities Have Not Adopted It
Five cities is not 19,000. There are roughly 19,000 incorporated municipalities in the United States, most of which still run the same manual process they used in 2005: PDF submitted, reviewer opens the file, checks it against the code section by section, sends corrections, waits for resubmission.
Three reasons explain the inertia, all institutional rather than technical. Procurement cycles designed for physical goods, not SaaS, mean that getting a city council to approve a $50,000-per-year software contract requires an RFP, a vendor evaluation, and political will from elected officials who may be nervous about “artificial intelligence” appearing in a budget line item during an election year. Liability remains unlitigated: if an AI misses a code violation that a human reviewer also misses, and a homeowner discovers it after occupancy, who is responsible, the city, the software vendor, or the reviewer who signed off on the AI’s output? And workforce politics loom large, because plan reviewers represented by public employee unions can see the arithmetic of a tool that does 75 percent of the work in 25 percent of the time, regardless of how carefully vendors label it “augmentation.”
What AI Cannot Check
AI permit review works well on objective, rules-based compliance: setbacks, height limits, lot coverage ratios, fire separation distances, energy code minimums. It does not work on subjective review categories like design review board compatibility, neighborhood character assessments, or variance determinations, which require judgment that integrates community context and political sensitivity in ways no current model handles reliably. A setback is either 15 feet or it is not; whether a proposed second-story addition is “compatible with the existing neighborhood character” depends on which neighbors show up to the planning commission meeting.
The split runs roughly 60/40 in most jurisdictions, with about 60 percent of a standard residential plan review involving objective code checks that AI handles today and 40 percent requiring experienced human judgment. CivCheck partially addresses the experience gap by capturing senior reviewers’ knowledge about local interpretation patterns into the platform, creating institutional memory that survives individual retirements and compresses the learning curve for new hires from years to months.
What This Means for You
If you are building a home in Honolulu, Austin, Bellevue, Naples, or Louisville, your application moves faster and your carrying costs drop. If you are building everywhere else, your permit sits in the same queue it sat in a decade ago, checked by a department that may be short-staffed because its most experienced reviewers retired, and the person reviewing sheet A-3 may miss the setback encroachment that adds six to eight weeks and $3,000 to $5,000 to your timeline.
The technology exists, it is deployed, and it produces measurable results. The question is whether the other 18,995 municipalities will adopt it before the last generation of experienced plan reviewers finishes cleaning out their desks.
Limitations of This Analysis
The ICC workforce data is from 2014 to 2017, and no comparable survey has been published since, meaning the 54 percent figure likely undercounts the current problem. CivCheck’s and Archistar’s performance claims are company-provided or based on city-reported data, with no independent third-party accuracy audit published. Carrying cost estimates vary by market: Austin’s $9,700 per month is high-cost; in cheaper regions, the dollar impact is smaller, though the proportional burden may be equivalent. Review time reductions apply to residential permits specifically; commercial and industrial permits involve complexity current AI tools do not fully address.