A Hawaii Story: An FBI Sting, Building Permits and AI

Two months after Dawn Takeuchi Apuna arrived at Honolulu’s Department of Planning and Permitting in early 2021, the FBI showed up.

On March 30, federal prosecutors charged five current or former department employees and an outside architect for bribery. They later pleaded guilty for illegally taking or giving money to speed up permitting and approvals – illustrating how a badly backlogged system can open the door to criminal behavior.

“When there’s backlogs, people take advantage,” said Takeuchi Apuna, now the department’s director, in an interview with The AI Innovator. “They say, ‘Oh, I can get you a permit faster.’”

The arrests added urgency to a modernization effort that would eventually include new permitting software, new operating procedures and AI. “If we can just get permits out faster, people can’t take advantage of the fact that we’re taking forever,” she said.

Honolulu piloted an AI system to make the permitting process faster, and it is now expanding its use. The city is using CivCheck, an AI-assisted plan review system that helps applicants and city employees identify incomplete information and non-compliance in applications. Residential applicants can voluntarily use it, but it will become mandatory in September with commercial projects slated to follow later this year, Takeuchi Apuna said.

Catching problems before the queue

Takeuchi Apuna inherited a department using permitting technology that was roughly 30 years old. But replacing the underlying software addressed only part of the bottleneck. Poor-quality applications could go through about half a dozen rounds of corrections, she said. This delays permit approvals.

CivCheck, an AI system from Clariti, checks building plans for missing information and possible code problems. It flags issues for applicants before they submit their applications for remodels, additions or new construction, saving the city and applicant a lot of time from having to return the application to be corrected. City employees still make the final decision of whether or not to issue the permit.

Using the new AI system, residential plan review time fell by 70% in the pilot, according to Clariti. The average time for staff to make a decision was cut in half: from 73 days to 32.5 days in the first quarter of 2026.

Takeuchi Apuna said the department used the pilot period to find gaps and bugs before rolling out the system more widely. Employees also had to be convinced that AI was intended to assist rather than eliminate them.

“This is not to replace your jobs, but to really just give you tools to help you do your job better,” she said.

Selling AI to City Hall

For AI companies hoping to sell to government, Honolulu’s experience also illustrates hurdles that have little to do with model performance.

Public agencies must contend with procurement rules, employee and public trust, limited staff and annual budget cycles. Vendors, Takeuchi Apuna said, need patience and a detailed understanding of the government process they are trying to improve.

“Just know that government is built differently. It’s about the public trust. So we’re very careful about what we do and how we do it,” she said. Vendors must be patient because “most jurisdictions are strapped − they don’t have the resources, they don’t have the funding, they don’t have extra people.”

Funding can be an even larger obstacle. Honolulu used federal pandemic recovery money to pay for CivCheck and its broader permitting-system modernization. Without federal money, she said, the city might not have been able to pursue the systems when it did.

That cushion is disappearing. The department now needs to make the case to the City Council for ongoing licensing and other technology spending, potentially passing on costs as plan-review fees. Demonstrating measurable returns matters particularly because Honolulu has struggled to hire engineers at government salaries, Takeuchi Apuna said.

Cities move cautiously into AI

Honolulu is part of a broader but uneven municipal experiment with AI. In 2024, an International City/County Management Association survey of 635 local-government practitioners found that 48% considered AI use a “low priority” and fewer than 6% called it a “high priority.” Only 10% had personnel overseeing AI efforts and 9% had an organization-wide AI policy.

Since then, momentum has picked up. Boston is testing AI search for residents and has adopted a citywide generative AI policy, while San Jose, California is using AI to fix residents’ pain points. For example, it rolled out a system that cut bus travel times by 20%, CIO Stephen Caines told The AI Innovator earlier this year.

This year, the National League of Cities reported that 96% of mayors are interested in using AI, even as many cities still lack the staffing and governance infrastructure to deploy it safely. The group has created the AI & Emerging Technology Forum to address AI use.

Honolulu is already looking beyond plan review. The permitting department is exploring the use of AI tools that could read property records in microfiche, paper and digital formats.

It would scan the documents, label and index them, potentially turning records that could take years to process manually into searchable digital files. The department is talking with vendors and is also considering technology for mobile inspections.

“People will be able to just go online, get all their information, and then be able to do what they need to do,” she said. “That’s just huge.”