How Local Governments Can Use Guardrails to Accelerate Planning and Permitting With AI

• AI is already embedded across local government processes, from vendor software and GIS tools to staff’s personal productivity apps and resident’s preferred tools.

• AI guardrails for local government can help agencies manage risk while still accelerating planning, permitting, inspections, and public service delivery.

• The best AI strategies start with real operational pain points, create staff capacity for higher-value work, and prepare departments for the next levels of agentic AI capabilities.

The annual Transforming Local Government (TLG) Conference dedicates its three days of programming and networking to helping public servants and governmental departments innovate and navigate change. However, in the case of AI, the innovation has already pervaded local governments. Now they have to figure out how to put the right guardrails in place to meet the public’s needs, according to Dustin Haisler, Chief AI Office and US General Manager of Darwin.

Haisler presented “Scaling AI Responsibly: Guardrails That Enable, Not Block, Innovation” at TLG 2026 in Dallas, Texas. In it, he noted that like it or not, AI is hitting local governments from all sides. Their constituents are using it; their vendors are embedding it into software, and their employees are using it, whether or not they’re supposed to be. With how pervasive AI is, he says governments have to figure out how to adopt it responsibly.

“It is an incredible time to be in government,” Haisler says. “There are interesting technologies and companies using AI to solve very tactical problems for you. Now you have more capabilities than you could ever imagine. You just have to pick the right ones.”

Responsible AI use for government agencies begins not with a ban, or a policy document, but by clearly understanding where AI is already touching governmental workflows. Haisler says that in his audits, about 62% of AI use in a department was unknown or not considered by the staff. For example, the popular ArcGIS uses AI; even calendar schedulers usually employ AI.

“It’s not just a standalone thing anymore,” Haisler says. “Every vendor out there is probably working on some type of AI in their product, even if their product is not an ‘AI product.’”

Staff members working on city planning, permitting, inspections, GIS, document management, or public engagement are usually just trying to be productive, not malicious, by using AI technology. They’re seeking ways to answer citizen questions faster, write clearer communications, interpret complex code and bylaws, and keep up with increasing demands. However, good governance needs visibility into what tools are being used, to avoid data breaches or compliance violations.

While Haisler mentions visibility into AI tools being used because “you can’t govern what you can’t see,” he also cautions against making guardrails against them so prohibitive that staff just does their AI work on their own devices. That could cause liability and risk issues.

“Think about guardrails not as a brake but as an accelerator,” he says. And remember that existing policies around acceptable use, data security, and open records laws at the state level already apply to AI use within local governments. So those act as a starting point to forming a policy and culture around AI use, which Haisler says there is a small window of time in which to do.

With that urgency in mind, Sarah Endres, Technology & Performance Analyst for the Town of Gilbert, Arizona, supplied the operating model in her presentation, “Gilbert Forward: Tech, Trust, and Transformation.” She described the three “Ts” of her talk as pillars that are only successful when the all support the same one mission.

“Tech: how we find, adopt, and deploy the right tools,” she says. “Trust: the governance framework that’s made responsible adoption possible, and transformation: the culture that has shifted the mindset to make our efforts scalable.”

This framework applies well to planning and building departments. Technology alone will not fix a permitting backlog. Governance alone will not create adoption, culture alone won’t protect sensitive data. All three need to support each other.

If the Town of Gilbert is to lead by example, its followers should let their work guide them to the technology they need. Endres says that their AI-related workflow innovations often come from discussing strategy and the direction they want the town to go, rather than specifically looking for tech tools.

“Some of the best opportunities we have didn’t stem from technology conversations,” she says. “You’re just hearing the frustrations of those around you.”

City planning and permitting teams experience those frustrations for example from the public’s confusion about process, applications, and submittals; and time spent on lengthy inter-departmental handoffs or searching through records, codes, and other info.

AI can ease those pain points if departments can clearly define their use cases and keep humans accountable for judgment. Matching AI tools with needs also matters. For instance, if a staff needs help approving applications, make sure a tool is not simply for summarizing applications. Choosing AI tools wisely to match the most dire needs can reduce administrative drag and free up staff for work that exercises their discretion and expertise and enhance public trust.

The quicker and more comprehensively a local government establishes its digital infrastructure and AI guardrails, the better prepared it will be for the coming next stages of AI’s capability.

“The scary stuff comes when the AI agents can operate autonomously or semi-autonomously,” Haisler says, “where you give it a goal, and it can self-assign tasks and continue to evolve.”

New risks could arise there around identity, accuracy, abuse, and intent, like if AI agents do things you didn’t expect or want in service of a goal that you defined. 

However, this type of more advanced agentic AI could also improve public access and reduce confusion by helping people research zoning, complete forms, respond to corrections, monitor permit status, schedule inspections, and so on.

As before, the next stages of AI will come with both positive and negative potential. Local governments can prepare by establishing clear standards for AI-assisted submissions. Rules and data should be machine-readable and systems should be auditable. Any information access, whether by humans or digital systems, should be transparent and accountable.

“Start to think about how AI agents will intersect with your workflows and what will be okay,” Haisler says. “We’re going to have to make sure our workforce is ready to be part of that.”

In addition to preparing for what could go wrong, another, perhaps happier, question to ask is “what could go right?” Endres framed it a different way: “Maybe the question you ask is not which AI tool we should adopt, but what can your organization achieve in that free time that AI creates for you?”

She continued that digital tools can create extra time for staff, but that efficiency is wasted if they immediately fill the time with more busy work. Endres compared the dilemma to household technologies like robot vacuums, dishwashers, or grocery delivery apps. If you don’t do something valuable with the time saved, then what do you gain?

“This represents the risk of AI adoption if we’re not intentional about it,” she said. “When we’re able to do more meaningful things, that’s when we see the impacts of AI. That will tell a more accurate story of the successes of implementing any emerging technology tool. That’s when you’ll start to see your culture shift.”

City planning and building departments can concentrate on freeing up time from AI to create capacity for higher-level work like giving permit applicants guidance earlier and more often, more consistent review comments, proactive community communication, and complex judgment calls.

AI is already changing how resident communities learn about zoning, prepare applications, and interact with government. Like it or not, departmental staff is also already using AI in some capacity. Local governments can’t ignore AI. And if they respond with only restrictions, they risk pushing AI use into personal accounts and unmanaged devices.

Instead, AI guardrails for local governments can build a culture of trust around AI use through the visibility of how residents, vendors, and staff interact with public systems.

Endres says that local governments almost universally ask the same question of where to begin with an AI policy. However, waiting for perfect clarity is not a strategy. “You don’t have to have it all figured out to start,” she says.

They can begin with practical entry points like mapping out where AI is already present in their workflows, classifying use cases by their risk vs. value, and clarifying which uses are approved and what is prohibited. Endres also suggests starting by piloting one high-friction workflow and measuring how much capacity it creates.

Fortunately, both Haisler’s and Endres’ messages pointed out that AI adoption is not a choice between innovation and control. Responsible AI guardrails for local governments establish both by building a foundation around clear policy, human review and auditability, and transparency. Done right, it will enable staff to provide higher-value public service.

“The question is not whether AI will change government,” Haisler says. “It is whether government will be ready when it does. You just have to build the right foundation, and you can accomplish anything.”