Cities turn to AI to speed housing permitting

This article was originally published by Stateline.

New housing projects can sometimes stall out as a developer applies to city or county planning boards and goes through cycles of inspections, reviews and approvals. 

Now, cities are increasingly turning to artificial intelligence to address one of the most persistent sources of housing permitting delays: incomplete applications. And two new streams of federal money aim to help them.

AI can handle time-intensive tasks such as scanning applications and flagging errors. The aim, some city officials say, is to eliminate repeated rounds of corrections that can add weeks or months to the process — but to leave final decisions in the hands of human reviewers.

The sweeping new federal housing law enacted last month creates an Innovation Fund with $200 million annually for local efforts to speed up permitting. And the U.S. Department of Housing and Urban Development is offering up to $3 million in grants for local governments to deploy automated building code permitting systems.

Syracuse, New York, applied for one of the HUD grants, along with cities across the country, including Coeur d’Alene, Idaho; Mobile, Alabama; and Richland, Washington. 

Vincent Scipione, chief information officer for Syracuse, told Stateline that if the city wins the grant, it would open up a competitive bidding process for vendors and ask them about two big topics: data governance and algorithm design. The city wants to install an AI system to assist developers to file more complete applications upfront thus reducing time needed for revisions.

“Who owns the data that is being handled, and how is it being handled? That’s important,” Scipione told Stateline. “If it ends up being an AI-based solution, then algorithms that are involved and we need to ask what they are going to be handling?” 

Large cities including Baltimore, Denver and Los Angeles have recently launched AI tools to streamline housing permitting, and smaller cities such as Everett, Washington; Lebanon, New Hampshire; and Naples, Florida, are following suit.  

Texas’ Harris County, which includes Houston, this month put $750,000 toward an artificial intelligence program meant to speed up the permitting process for new construction projects.

The Denver City Council voted in March to approve a five-year contract for an AI Guided Plan Review platform to reduce permitting backlogs, for around $4.6 million. A city official said only about 37% of applications are accepted on the first round, and the goal is to increase that to 80% with the new tool. 

The lone dissenting vote — council member Flor Alvidrez — raised concerns about the inaccuracies of AI systems generally. She also said local architects and developers raised concerns with her that AI may not be ready to handle the complexity of zoning codes and site-specific conditions. 

Matt Mudd, press secretary for Louisville, Kentucky, Mayor Craig Greenberg, told Stateline the city is testing an AI-assisted residential permitting system aimed at reducing delays caused by incomplete applications. The pilot uses property information, Geographic Information System (GIS) data and local permitting requirements to flag missing or incomplete information before an application reaches a reviewer, with early testing capable of reducing “avoidable” resubmissions by 50% or more, Mudd said.

The project is in the testing phase with a small group of builders and has not yet launched publicly. Mudd said employees remain responsible for permit reviews and final decisions.

CivCheck

Honolulu was one of the first cities to introduce AI into its permitting process with a tool called CivCheck. That program examines building permit applications to check applications against local codes and bylaws, flag missing information and guide applications through corrections. 

Dawn Takeuchi Apuna, director of the city and county of Honolulu’s Department of Planning and Permitting, said the better the quality and completeness of the applications, the better and faster the department’s review.

Prior to agreeing to work with CivCheck, Takeuchi Apuna said, city officials made sure the vendor knew that they weren’t looking to replace staff. She wanted AI to replace time spent by city planners reading incomplete applications and sending them back, and free them up for more face-to-face meetings with developers.

The rollout of Honolulu’s overhaul of its decades-old permitting process, which included a cloud-based platform called HNL Build as well as CivCheck, wasn’t entirely smooth, however. An anonymous survey of more than 150 employees — half the staff — in the weeks after HNL Build launched provided almost universally negative feedback and many urged  a return to the old system, the Honolulu Civil Beat reported.  

The median wait time to obtain a new building permit dropped to 2.5 months from August through April — a 40% decrease compared with the same period a year earlier, a Civil Beat review found.

So far, Honolulu has reported that CivCheck pre-screening has dropped permit reviews from an average of 3.4 review cycles per application to 1.4 cycles for single- and two-family residential projects. A single review cycle is the time spent by a planner examining a permit application, identifying problems or incomplete information, sending it back to the applicant for corrections, and then reviewing the revised application.

CivCheck applications averaged 7.7 corrections, compared with 23.5 corrections for applications that did not use CivCheck, and the average time an applicant spent going through the city’s permitting process decreased from 73 days to 32.5 days, both officials from the city and county of Honolulu and CivCheck told Stateline.

“We want to get back to just the face-to-face and the more complex part of the review,” Takeuchi Apuna said. 

She added that AI “doesn’t necessarily catch or know how to make the decision on” the grayer areas of a project — where a proposal might meet basic requirements on paper but have unusual site conditions or neighborhood impacts — while a planner knows the context and ambiguities to determine how a rule or code applies.  

Seattle also crunched the numbers during its pilot of CivCheck and found it was 87% accurate on application completeness checks and 92% accurate on design-compliance checks. CivCheck showed an approximate 50% reduction in average days for intake review of permits and a 35% reduction in correction cycles per review.

However, Seattle officials noted in their reports that while CivCheck showed promise, the city was not ready to adopt it fully. As of its June 2026 report, the city was still exploring funding and logistical support for tools such as CivCheck. 

CivCheck told Stateline it was working with more than 20 cities, ranging from pilot projects to full implementation. 

“When you can remove three, four or five review cycles of feedback and turn that into one review cycle, that’s where you start to see material improvements in permit turnaround times,” said Cyrus Symoom, co-CEO of Clariti Software, a community development and permitting solutions provider that acquired CivCheck. 

Their tool, Symoom said, is solely to be used as a fact-checker but isn’t meant to replace or take the place of human decision making on permits. Ultimately, if a mistake is made, it’s up to the planner to properly vet that information.

“If you’re just gonna take the existing (permitting) process and bring it digital without an idea of what part of the process you want to fix, you’re not gonna see any real change,” he said. 

A Two-Way Street

Zhenia Dulko, a planner and researcher with the educational and membership organization American Planning Association, said that group has identified more than 70 state and local cases of AI being used across planning functions, including 13 for zoning and development. 

While automation can take technical and onerous work off planners’ plates, Dulko believes  it allows planners more time for the work AI cannot perform such as talking to residents, connecting stakeholders to projects and front-facing public engagement. 

“AI will help speed up some of this work but there’s a lot of this planning work that doesn’t need AI because humans want to talk to humans about their ideas and visions, and find a way to finish that project.”

Dulko said planners also consider factors such as the surrounding community, inclusion and equity goals that may be harder for an automated system to replicate.

“Efficiency in terms of numbers is important but in terms of time it’s even more important,” Dulko said. “But we just should not miss these very other important things. We should remember that this is just one part of the permitting equation.” 

An August 2026 Journal of Urban Economics paper looked at Los Angeles data on development timelines for all multifamily housing projects permitted by the city between 2010 and 2022. The analysis estimated that reducing approval times by 25% could have increased the rate of housing production by almost 24%, if the researchers accounted for both already-started projects that would have been finished sooner and the effect of incentivizing new development. 

But some city officials caution that while AI can help speed the permitting process, applicants are responsible for some of the delays.

“I think a lot of people assume that it’s all on the local permitting agencies, and think we’re this big black box that no one can figure out what we’re doing and that we’re just sitting on applications,” said Takeuchi Apuna. “It’s important to educate people and let them know that the applicant plays a major role. CivCheck helps address that role, but permitting is really a back-and-forth process. Applicants account for at least 50% of that whole dance and the actual review time.”

Saf Rabah, CEO and founder of Govstream.ai, an AI tool that has contracted with Louisville and Bellevue, Washington, says the permitting process is a two-way street that relies on both city and developer. He said the process could use more guidance from cities to help applicants who may not be well-versed with building permitting.

“A lot of people end up applying for the wrong permit and go down this rabbit hole when they upload their plans. There is no validation. There is no guidance. There is literally nothing. It is a minefield of ‘gotchas’ in front of a builder,” said Rabah.