Beyond 10 Seconds: Why AI's Real Promise for Building Permits Is Risk Prediction, Not Speed
Key Takeaways
- ▸The real problem with building permits isn't approval speed but unpredictability—the same permit type in the same city can vary by 8x depending on which examiner reviews it, queue load, and other opaque factors
- ▸Permit timeline variance creates millions in carrying cost risk for developers; a $25M project could face a $2.3M swing in financing costs based on P50 vs P90 outcomes
- ▸AI's highest value is providing predictive risk intelligence on permit timelines before developers close on sites, not automating yes/no approval decisions
Summary
Jeff Bezos recently went viral advocating for AI to approve building permits in 10 seconds, but new analysis of 1.8 million permits across seven major U.S. cities reveals the real problem isn't approval speed—it's unpredictability. Data shows massive variance in permit timelines: Austin's median is 22 days but the 90th percentile is 179 days (an 8x difference), while NYC ranges from 79 to 344 days. This unpredictability creates enormous carrying cost risks for developers; a $25M San Francisco project could face a $2.3M difference in financing costs depending on whether the permit approval falls at the median or 90th percentile. Prevesta, which has built a database of 1.8 million building permits and 1.5 million duration records across major cities, argues that AI's value lies not in automating approvals but in providing risk intelligence that lets developers understand their realistic timeline distributions before closing on sites or drawing down construction loans.
- Cities lack standardized, machine-readable permit data and most can't query their own historical approval rates by permit type, examiner, or time period
Editorial Opinion
Bezos's 10-second approval sound bite captures public frustration with sluggish government but misses the deeper economics that actually constrain development. The Prevesta analysis reveals that permit variance—not approval speed—is what kills projects and allocation decisions. If AI can help developers price timeline risk accurately before committing capital, that's far more valuable than shaving approval time from months to seconds.
