According to HousingWire, cities across the country are discovering that artificial intelligence can meaningfully speed up the permitting process. Honolulu has already cut its permit review times in half by adopting AI tools, showing that the technology has real potential to unlock faster housing approvals. However, experts discussing this issue at a Bipartisan Policy Center event made clear that AI's ability to deliver results depends heavily on the quality of the data feeding those systems.
The core challenge is that zoning and building code information in most jurisdictions is outdated, inconsistent, or just plain messy. When AI encounters contradictory rules, missing information, or exceptions scattered throughout hundreds of pages of code, it tends to make mistakes that actually require more human review, not less. This defeats the purpose of speeding things up. One expert pointed out that the real culprit behind slow permitting is not usually one major roadblock, but rather a thousand small inefficiencies piling up. She noted that only about ten to fifteen percent of applications arrive complete when first submitted, which means municipal staff spend enormous amounts of time asking for missing information and coordinating back and forth with builders.
The underlying problem runs even deeper. Developers often struggle just to figure out what they are legally allowed to build on a piece of land, and different cities use different terminology for the same zoning concepts. One researcher found that about a third of jurisdictions have conflicts between their written zoning text and their zoning maps, and some districts appear on maps but no longer exist in the official code. When the information is this inconsistent, even sophisticated AI tools cannot reliably guide the process.
One homebuilder described a project where his company built homes that technically complied with general zoning rules, but a separate regulation about facade design and entry placement was overlooked initially. The AI missed it, and the city only flagged the issue after state approval had already been granted and manufacturing was underway. This real-world example shows how easily hidden zoning requirements can derail projects, even with AI assistance.
What I am seeing locally in the Bay Area and particularly in the East Bay is that we have the same fundamental problem these experts are describing. Our jurisdictions have accumulated decades of codes, variances, and overlapping requirements that even experienced professionals sometimes struggle to navigate. Until we invest in cleaning up and standardizing our zoning data at the municipal level, technology alone will not solve our permitting bottlenecks. The real opportunity here is combining clearer codes, better data, and yes, AI tools too, so that what should be a straightforward process actually moves forward without constant delays and requests for clarification.
