Chad Catalino, chief public defender in Allegan County, Mich., had managed to snag a recent graduate from a prestigious law school for his office. She’d found her first private-sector job exhausting and unrewarding. She wanted to make a difference, and his office provided a chance to do that. But after she’d had time to take on her new responsibilities in full, he found her crying at her desk. She told him there was no way she could review the thousands of hours of body cam footage necessary for her cases — at least not if she wanted to have time for any other parts of her job.
As body cam, dashcam and surveillance footage have become more common, public defenders have struggled to keep up with the deluge, especially as caseloads have remained stubbornly high, Catalino says. The audio and video evidence that accompanies criminal cases has increased as much as 600 percent; one felony case could come with 50 or 60 recordings that need to be examined to determine what really took place on both sides of an arrest.
AI has managed to ease this burden for Catalino’s attorneys. The tool can search hours of video to find — and transcribe — sections relevant to a case. An attorney can ask to see footage of a suspect at the moment of arrest, for example. What this means, Catalino says, is that well-paid attorneys can spend more time focused on the legal issues in the case. Taxpayer dollars aren’t being spent on work that doesn’t take advantage of their skills. Case clearance is up 8 to 9 percent after about a year of implementation.
The tool Catalino is using began as a project of two University of Michigan computer science students. Seven statewide systems are now using it, including those in Montana, Iowa and Kentucky. Devshi Mehrotra, co-founder of Justice Text, interned at Google, Facebook and Microsoft while she was in college. “We learned things in those incredibly fast-paced, innovative environments,” she says. “But we wanted to do something purposeful.”
For Catalino, the tool has been a boon. His attorneys’ increased capacity to mine video evidence has enabled the office to make better deals for their clients, he says.
Justice Text’s tool is just one of many ways that artificial intelligence is already changing how government does its work. AI is impacting everything from government contracts to permitting processes. Many state and local leaders are excited about the technology’s potential to lighten the administrative load; there are also those that fear overreliance on the technology could lead to crucial errors in important work, or could cause serious job losses across the public sector.
For the bright young attorney in Catalino’s office, the tool has been a lifesaver — she’s managed to get on top of her caseload.
AI and the Public Workforce
Computer scientist Roman Yampolskiy’s warning that AI might leave 99 percent of workers without jobs by 2030 exemplifies rhetoric creating unease about the impact of a technology that investors view as the hope of the American economy. Tim Ewell, chief assistant administrator for Contra Costa County, Calif., doesn’t expect AI to replace many of the humans that work for him. “Our approach has very much been that this is a tool for people to be more productive in the public sector,” he says. “We still require a human in the loop.”
Ewell points to the 5,000 or so contracts that the county produces each year for its many vendors. AI can draw on this library to create new contracts or contract extensions that get most of the way to completion. Human specialists can then bring them to final form.
Micah Gaudet, a deputy city manager in Maricopa, Ariz., sees big value in low-cost AI tools that can help finance workers debug spreadsheet problems. These help them untangle technology issues more effectively than video tutorials they might otherwise rely on, and get them back to focusing on budget analysis. Contra Costa has deployed Microsoft Copilot throughout the county and offers biweekly training sessions to help workers understand its capabilities.
A recent analysis by Darwin, a private company focused on assisting governments with IT implementation, found that employee use of standalone tools such as Copilot and ChatGPT accounts for the majority (62 percent) of IT use in public-sector agencies. “The challenge is no longer adoption — it is governance,” it concludes.
AI implementation is an exercise in change management, Ewell says, an inflection point comparable to the arrival of the personal computer. “The people in the organization that try to understand and integrate AI as a tool are going to become so much more productive that they will be the people who rise up in the organization.”
This strategy may not be enough for all workers. In a paper published this year by GovAI, a research organization focused on AI governance and public policy, senior research fellow Sam Manning identifies occupations at most risk of being replaced by AI systems. In his estimation, jobs including clerks, administrative assistants, tax collectors, interviewers and receptionists have high degrees of exposure and low capacity to adapt to changing technology.
AI is proliferating across local governments for all kinds of uses.
Quicker and easier translation is one of them. The U.S. is the most linguistically diverse country in the world — Americans speak well over 300 languages at home.
Wordly
AI for Translation
Since 2023, North Las Vegas has provided real-time AI translation of its public meetings, the first jurisdiction in Nevada to do so. About 40 percent of the population speaks a language other than English, City Clerk Jackie Rogers says. Spanish and Tagalog are most common. At these meetings, large TV screens display translations of remarks by officials and attendees as they occur. Spanish is the default, but attendees can scan a QR code and access translation in virtually any language through their phones. Audio translation is included, to offset any literacy issues. Residents attending meetings can come to the podium and address city officials in their native language, and their remarks will be translated into English for officials.
Some of the first such exchanges involved community members approaching the podium to express excitement that the system was in place, Rogers says. Resident Maria Navarette, a member of a local organization that advocates for marginalized groups, thanked the City Council in Spanish. “La tecnología de traducción es algo más que una comodidad, es una herramienta vital para gente como yo. (Translation technology is more than just a convenience, it’s a vital tool for people like me),” she said.
Such an exchange would be equally seamless in Ukrainian, Rogers says. “There’s almost no language that you can’t come to City Hall and speak.” Dozens of cities have put similar systems in place, particularly in California and Nevada. Contra Costa County has implemented one for its board of supervisor meetings (a use case in which human workers have been removed from the equation). “We no longer have translators and interpreters,” Ewell says.
AI for Prescription Renewal
Utah made headlines at the beginning of 2026 when it became the first state to allow patients to refill prescriptions online through an AI-driven platform. The Office of AI Policy — which launched in July 2024 — oversees the service and aims to balance excitement about AI with public anxiety about potential pitfalls and consequences, says Margaret Busse, executive director of the Utah Department of Commerce.
In Utah, only a licensed doctor can write a prescription. But the Office of AI Policy and the Department of Commerce opted to allow a trial of AI-powered renewal. The office created guardrails around the process — it only applies to prescriptions that are 30, 60 or 90 days out, for example. The initial prescription must always have been made by a physician. And refills for drugs that are addictive, like opioids, are not allowed.
As a further safeguard, the deal with Doctronic was structured to require that a physician licensed in Utah would have to sign off on 250 interactions on the platform before AI took complete control, so the state could watch for problems. The service was introduced in January as a pilot project with an initial year of regulatory relief and the possibility of a second year.
The pilot has significance beyond proof of concept for an AI application. Almost half of all patients don’t take medications as prescribed, and about a third don’t fill even their first prescription. This behavior is associated with hundreds of thousands of preventable deaths and hospitalizations. Busse has heard from physicians that they don’t necessarily have the bandwidth to pay careful attention to renewals; she’s interested to see if renewal using AI could improve due diligence. The platform is expected to benefit rural residents with less ready access to in-person medical care. The first stage is to watch, oversee and observe, Busse says: “After this period of temporary regulation has concluded, our office can decide whether or not to make the recommendation to the Legislature to change the regulation.”
It’s still early days for the pilot and Utah’s overall approach with AI is to proceed cautiously. “The idea is to make sure that we are harnessing AI that’s actually going to be helpful to people,” Busse says. Her department has sued Meta, TikTok and Snap for practices that it believes harm and manipulate Utah children. Gov. Spencer Cox is similarly concerned about the possibility of negative impacts from other AI products. “We’re working on a bill right now on AI companions and also one on deepfakes,” Busse says. The prescription refill pilot aims in the opposite direction. Busse hopes this use of AI can help create a path to better health-care access and lower costs.
City of North Las Vegas
AI for Permitting
In January of this year, the nonprofit Americans for Prosperity issued a report titled A Guide to Unlock Energy Abundance: The Impact of Permitting Delays on Access to Energy and Economic Development. According to their analysis of 30 proposed projects in six states — Arizona, Montana, Nevada, Ohio, Pennsylvania and West Virginia — dysfunctional permitting systems led to the loss of “at least 50,000 new jobs and nearly $75 billion in economic benefits.” Some projects had been delayed for as long as 15 years.
Permitting and regulatory red tape is an issue for both developers and local governments, adding “hidden taxes” to projects, hindering housing and business development and slowing economic growth, says Stuart Lacey, a regulatory technology and generative AI expert who holds 13 patents. Lacey says AI has only recently evolved to the point that a tool capable of transforming the permitting process was possible. The resource he developed, Labrynth, reviews draft applications from permit seekers and lets them know what they need to address to meet regulatory requirements.
The permitting platform launched in July 2025. In September, Lancaster, Calif., was one of the first cities in the country to make it available to permit seekers. City experts still review and approve applications, but pre-submission review means that when an application gets to them, there are fewer bumps in the road to approval. “Governments shouldn’t be adding to the cost of things,” says Lancaster Mayor Rex Parris. “They should be making them cheaper, more affordable, and AI really helps us do that.”
Lacey puts the scale of possible hurdles in perspective. As of January 2026, building permit applications must comply with changes to the California building code enacted in 2025. “We’re talking about 18,000 pages, 21 million lines,” he says. “A very cumbersome, extraordinarily difficult set of requirements.”
Permit seekers in Lancaster can upload their applications to Labrynth’s permit tool and receive a score letting them know how well their application adheres to the updated building code.
It’s early to have much feedback from developers about their experience with this new service. “But what the staff is telling me is that we’re moving much faster, and that’s what the goal was,” Parris says.
Parris, a Republican, has pushed the envelope in multiple directions during his 17 years as mayor. In 2014 he set the goal of making Lancaster the first net-zero energy city in the country, starting by mandating solar energy for every new home built there. By 2019 it had achieved this goal, generating more energy from solar panels than it consumed. In recent years, it’s worked to bring in renewable hydrogen projects. Chinese manufacturer BYD built a half-million-square-foot facility in Lancaster, where it makes electric buses.
He’s equally bullish on AI. In Lancaster, police officers use the technology to create draft reports after an arrest. Cameras outfitted on city maintenance trucks use AI to detect code violations and send letters requesting compliance. (These types of cameras add a new dimension to long-standing privacy concerns regarding video surveillance in public spaces.)
Parris agrees with his government colleagues that AI isn’t a substitute for human judgment; Lacey, the technologist, doesn’t expect AI to rule the world. “I deeply believe that the combination of both AI and humans in the loop is the best way to deploy smart AI,” he says.
