John Lorinc is a freelance journalist specializing in cities, climate, and technology. He can be reached at lorinc@rogers.com. 

The turnaround time was astonishing — a “game changer,” as one city manager put it.

When Bakersfield officials installed an artificial intelligence-based permitting platform last spring, the processing times for four categories of projects — rooftop solar installations, electrical panel upgrades, residential electric vehicle chargers, and re-roofing jobs — plunged from hours and days to moments.

“Instead of having a staff member spending 30 minutes per review to an hour,” says Phil Burns, development services director, “it’s instantaneous. For the customer, the difference is, instead of waiting a day or even 48 hours, again, it’s instantaneous.” And, he adds, “there’s no human error.”

Like all California cities, Bakersfield officials had to streamline rooftop solar permit applications to comply with a 2022 state regulation that made California in general, and Bakersfield specifically, an early adopter of AI-based permitting. In early 2025, the San Joaquin Valley city soft-launched an instant permitting platform created by Symbium. After asking a couple of rooftop solar installers to test the platform and identify bugs, Burns’ team rolled out the full service in July.

There’s a small irony in Bakersfield’s move. After the city in 2023 shifted from paper applications to digital forms, its average processing time for rooftop solar projects actually increased because staff were spending more time learning the new technology and doing data entry. The Symbium platform, Burns says, has reversed that trend, with staff now spending their time on more complex permits. “You look at those metrics and the staff savings time [and] it’s huge.”

“If we can reduce the friction there, ultimately we hope that we can get more shovels in the ground”

As California cities experiment with a range of AI technologies, automated permitting is proving to be one of the more promising platforms — one focused on a core municipal service long plagued by bottlenecks and red tape. Cities that have adopted AI permitting for fairly simple projects say that it speeds up processing and enables planning staff to dedicate their time to more complicated applications, which technology cannot — as of yet —handle alone.

The drivers of this trend include a bevy of state regulations, policies, and incentive programs enacted in the past decade or so that aim to streamline and accelerate housing development. “We’ve tightened up those timelines, which we call shot clocks, for approving entitlements, but also for issuing permits as well,” says Michael Lane, state policy director for SPUR (San Francisco and Bay Area Planning and Urban Research Association). “The only way the staff is going to be able to keep up is to really utilize these technologies.”

He points to San José, which began piloting an AI-based building permit approval platform developed by CivCheck last fall. “[P]art of the big delays and the issues that makes the housing crisis so bad in California is how long it takes,” Tasha Dean, a city spokesperson, told StateScoop when it launched. “If we can reduce the friction there, ultimately we hope that we can get more shovels in the ground.”

San José’s platform is designed to approve accessory dwelling units (ADUs). These backyard cottages now comprise some 40% of the city’s building permit applications. The benefits don’t come just from cost savings and faster processing, Lane adds, but also from increased predictability. “In the past, you were there fighting for two years to get the ability to even build an ADU, and the state really pre-empted that and took that discretion away,” Lane says.

While many permitting software firms are based in California, interest in the technology is not limited to the state. Cities in British Columbia, Alberta, and Australia have all adopted various AI permitting systems, but the motivations don’t vary greatly from place to place.

“For a very long time, planning has been seen as slowing down development because of a lot of rules, a lot of regulations,” observes researcher Lauren Andres, who holds a chair in Planning and Urban Transformations at the Bartlett School of Planning, University College London. “Obviously, you have this sort of contrast with the context and the complexity of a planning system, and a national narrative of pushing for more housing. So, you have AI seen as a big opportunity to try to shake things up with automation and make things much more efficient.”

The technology seems to dovetail with a range of local, regional, and national housing targets, such as California Governor Gavin Newsom’s target of 2.3 million homes by 2030 and U.K. Prime Minister Keir Starmer’s goal of 1.5 million new homes by the end of the current session of Parliament. “We’ve never been able to build enough,” says Andres of the U.K. housing system. “They’re really seeing this as a potential to try to do things differently.”

Approving a building permit might seem relatively straightforward. But such tasks, as planners and code examiners well know, are complex and involve layers of disparate and sometimes contradictory sets of rules — from building codes to local zoning ordinances and statewide planning laws — that defy automation. One basic issue is that even digitized blueprints and working drawings aren’t necessarily machine-readable. And the reams of building regulations and policies that exist in every locality often have yet to be translated into mathematical formulas that can calculate whether a given application is code-compliant.

Several firms now offer software platforms that purport to tackle these technical riddles, including CiviCheck and GovStream.ai. Symbium came at the problem via Stanford University’s AI lab, where co-founder Leila Banijamali, a technology lawyer, launched a project to translate laws into computer code to produce deterministic outputs, something currently not possible with predictive AI models. The question she and her collaborators posed: “How do we democratize access to regulatory analysis for the public?” The team decided to focus on other forms of government regulations, such as building and planning.

“It’s the core of every city’s activity, so there was a need for it,” she says. “When we came in, there was a bunch of mandates that were passed [for] accelerated approvals of permits for certain project types like accessory dwelling units, solar decarbonization projects, and [we] really jumped on those to help cities comply with those mandates.”

Another AI permitting firm, Archistar, approached this challenge from the vantage point of designers and planners. Founded in 2010 by an Australian architect, Ben Coorey, the company originally developed software that could generate compliant digital building concepts. According to Coorey, who holds a doctorate in generative design, architects who used generative design software needed a way to automatically evaluate whether those designs were code-compliant.

“We went through probably 10 iterations over five years of building an engine to store standardized zoning rules that would bring together all of these cities into a format that we could use to design buildings,” he says. “That was no mean feat because every city is unique in their zoning.”

Archistar’s early clients were in Australia, but the company has expanded its reach into the U.S., U.K., and Canada. Coorey says it’s working with Los Angeles contractors involved in rebuilding the communities destroyed last year by wildfires. The platform can automatically assess whether the plans for a proposed dwelling follow local zoning ordinances and building codes. Most homeowners, he observes, have never had to deal with the kinds of approvals needed to construct an entire house. “We were brought in to bring in our AI pre-checks to this process to help guide that process. Before you submit [plans to the city], you could check it with us.”

The company last November established a partnership with Vancouver designer Arno Matis Architecture and Urbanism to further refine its platform, with an eye to “accelerating multi-unit residential project approvals” in response to the city’s housing crisis. But whether these new technologies can meaningfully speed up approvals for projects more complex than rooftop solar panels is still an open question.

But for many city leaders, this much-heralded acceleration is a work-in-progress. Patti Garibay, community development director for Lancaster, California, says her team has a small pilot project that uses AI to check submitted project documents for compliance with local building and zoning codes. “The intent is basically to speed up the process because those are basic things that everyone should be meeting,” she says. “We’re just looking at kind of a small sample and trying the AI and building it out.”

The degree of automation in the permitting process also raises concerns about accuracy and the need for human oversight. “Once in a while, a planner might catch something, or the applicant might question something,” observes SPUR’s Michael Lane. “But I think [the AI platforms are] probably 80 to 90% accurate in the way that they can move forward. And,” he adds, “human error was also an issue in the past.”

But Archistar’s Ben Coorey acknowledges that some elements of more complicated permit applications will require human judgment. “Anything that is objective, measurable, quantitative — these systems can check it. Once you get to subjective and qualitative, at the moment, we’re thinking that’s still best left to the planners and the officials to review. Not all rules are black and white, and some rules are guidelines.” 

For now, Symbium and Bakersfield officials are developing a version of the firm’s software that can automate regulatory compliance approvals for ADUs, which are more complex in terms of design but can be built from generic plans provided by the city, and thus don’t demand qualitative judgments from planning staff.

“We’re seeing [ADU applications] take almost three months to go through planning and building, to get your permits, and some of them up to four months,” Phil Burns says. “We’d like to get that down to a month, but if we can get these pre-planned ones closer to instantaneous, that would really spur more use of this product.” As he adds, “Housing is really the underlying goal.”