Honolulu's building department deployed an AI plan-review system, CivCheck, and cut the median permit review time from 73 days to about 32, roughly a 40% reduction, while dropping the average number of review cycles from 3.4 to 1.4. It's one of the clearest efficiency wins from government AI adoption anywhere. And the humans in the department still review and approve every decision the system makes.

Building inspectors are rated Moderate risk, with a horizon toward 2043. The technology aimed at this job is advancing fast on the document-processing side, plan review, drone-based site inspection, automated code checking, and much slower on the parts that involve physically examining a structure and taking legal responsibility for the result. Like the skilled trades AI Doomsday has covered, the physical and the legal are the moat.

Underneath all of it is a workforce problem that AI is not positioned to fix. The International Code Council projects that more than half the current code-official workforce will retire within a decade. An estimated 40% are already over 55. The U.S. Bureau of Labor Statistics projects 0% employment growth for the occupation through 2035, which means every one of the roughly 14,800 annual openings is a replacement for someone leaving, and the pipeline of new entrants is not keeping up.

Key Points

  • Building inspectors are rated Moderate risk with a horizon toward 2043, as AI compresses plan review and site documentation while the physical inspection and legal sign-off resist automation.
  • Honolulu's CivCheck AI plan-review system cut median permit review time from 73 days to about 32, with humans still approving every decision; similar systems are live in Austin, Los Angeles, and Denver.
  • Drone and computer-vision platforms including DroneDeploy, Skydio, and Zeitview now automate progress tracking and visible-defect detection, but cannot tap, probe, or test a structure, and are weather-dependent.
  • The occupancy certificate is a legal instrument backed by a licensed inspector's professional liability and indemnity insurance; no jurisdiction allows an AI system to issue it.
  • The U.S. Bureau of Labor Statistics projects 0% employment growth through 2035, with the International Code Council estimating more than half the current workforce will retire within ten years, making the near-term problem a shortage rather than displacement.

What a Building Inspector Actually Does

The job has two phases. Before construction, inspectors and plan reviewers check that submitted drawings comply with building, fire, electrical, plumbing, mechanical, and accessibility codes. During and after construction, they visit the site at defined stages, foundation, framing, rough-in, final, to verify the work matches the approved plans and meets code, including work that will soon be hidden behind drywall or underground. At the end, the inspector signs off on a certificate of occupancy, a legal statement that the building is safe to use.

What AI Is Already Doing

Plan review is where automation has moved fastest. CivCheck in Honolulu and Denver, Archistar in Austin and Los Angeles, and UpCodes' AI-native plan review, which the company says covers 11 million code sections across thousands of jurisdictions, all pre-screen submitted drawings against code, flagging likely violations for a human reviewer to confirm. Cities report first-submission approval rates and review-cycle counts improving substantially where these are deployed.

On site, drone and computer-vision platforms, DroneDeploy, Skydio's X10, Zeitview, are automating progress tracking and visible-defect detection: cracks, corrosion, roof and facade anomalies, thermal issues. Reality-capture tools like Matterport and NavVis produce LiDAR-based digital twins that let some inspection and condition assessment happen remotely, cutting large infrastructure inspection timelines from months to weeks.

THE RETIREMENT CLIFF

Industry surveys have found roughly 80% of code officials expecting to retire by the end of the decade, against a workforce whose median age is around 46, several years above the national average. With BLS projecting flat employment, the occupation is not growing or shrinking so much as struggling to replace itself. AI plan review is being adopted, in city after city, specifically because there are not enough people to do the reviews, not because the people are being removed.

The Line at the Occupancy Certificate

The legal structure of this job is what caps the automation risk. Building inspectors are licensed by state, carry mandatory continuing education, and hold professional liability insurance because they are personally and professionally accountable for defects they should have caught. The certificate of occupancy is a legal instrument of that accountability. No jurisdiction has proposed letting an algorithm issue it, and the liability question, who is responsible when an AI-approved building fails, has no current answer that would let a city sign off on removing the human.

The technical limits reinforce the legal ones. A 2026 systematic review of automated code-compliance systems found that building codes' inherent ambiguity, nested legal and technical language, unit conversions, jurisdiction-specific exceptions, remains genuinely hard for language models, which have demonstrably failed on compliance reasoning tasks. Drones can photograph a roof but cannot walk it, tap a beam, or open a panel, and the defects that matter most in a structural failure are often the ones behind a surface a camera can't see.

The Self-Certification Experiment

The more immediate structural change to this job may not be AI at all. After the 2025 wildfires, Los Angeles issued an executive order letting licensed architects self-certify code compliance on reconstruction projects, bypassing municipal plan check to speed rebuilding. Similar self-certification programs exist in New York, Chicago, Phoenix, and elsewhere. This doesn't automate the inspector's judgment, it transfers it, along with the liability, to a different licensed professional. It's a response to the same shortage AI is being deployed against, and it reshapes the profession without a single algorithm involved.

The Labor Market Reality

The occupation covered about 153,900 jobs in 2025 with flat projected growth. Permit delays carry real economic cost, roughly 1% added to project cost per month of delay, which is the pressure driving cities toward AI plan review and self-certification alike. But the direction of that pressure is toward getting more throughput from a shrinking workforce, not toward eliminating the workforce. An individual building inspector's near-term risk is closer to being overloaded than to being replaced.

How to Use AI as a Building Inspector Now

For plan review: AI pre-screening is a genuine force multiplier for an understaffed department. Treat flagged items as a prioritized worklist, not a verdict, since the code-interpretation error rate is real.

For site work: drone and reality-capture tools are strong for progress documentation, visible-defect triage, and hard-to-access areas. They extend an inspector's reach; they don't replace the visit where something has to be physically tested.

For career positioning: the shortage means this is a field actively recruiting, and the inspectors who pair code expertise with fluency in the new plan-review and drone tools will be the ones running larger jurisdictions with the same headcount.

What I Think

The 2043 horizon looks reasonable, and it's the legal accountability, not the technology, that sets it. Plan review will keep getting more automated, and that's a good thing for a system where permit backlogs add measurable cost to every project. But the certificate of occupancy is a signature with liability attached, and until someone builds a legal framework for an AI to carry that liability, which no one is seriously attempting, the inspector stays in the loop.

The real story for this profession is the one the AI adoption is a symptom of: a workforce retiring faster than it can be replaced, cities reaching for automation and self-certification because they have no other option. AI isn't the threat here. It's the patch on a shortage that was going to force a change one way or another.

"The AI can read the plans faster than any human. It still can't be the name on the certificate that says the building won't fall down, and that name is the job."