Local governments are facing a pivotal moment. As veteran public servants near retirement and vacancies reach record highs, cities and counties across the country are grappling with how to preserve decades of hard-earned institutional knowledge. The challenge is clear: how do you capture the expertise of your most experienced employees while empowering new staff to hit the ground running?

During the 2025 ICMA Annual Conference session “Protecting Institutional Knowledge: How AI Can Transform Onboarding, Efficiency, and Retirement in Local Government,” panelists Erica Olsen of Madison AI, Peter Pirnejad, city manager, Los Altos Hills, California, and Dave Solaro, assistant county manager, Washoe County, Nevada, explored how artificial intelligence—specifically, custom AI models trained on local government data—can help solve this problem.

Accelerating Onboarding: Turning Information Overload into Institutional Access

Every new hire in local government faces the same uphill battle: navigating a mountain of records, policies, and processes that have accumulated over decades. Historically, much of this knowledge existed in the minds of long-serving clerks, planners, or managers—the people everyone went to when they needed to know “how things have always been done.” But as those people retire, that wealth of context risks disappearing. That’s where AI is changing the game.

Los Altos Hills city manager Peter Pirnejad described how his small, 30-person organization used AI to bridge this gap. He explained, “When our longtime city clerk retired, we lost 20 years of institutional memory overnight.” By uploading the town’s resolutions, ordinances, staff reports, and meeting minutes into a custom AI model, the team created a searchable “memory bank” that new staff could access instantly.

Now, instead of asking, “When did we last have a contested mayoral rotation?” or “Why was this policy adopted?” staff can query the AI and receive not only the answer but also citations back to the original public record. 

“It’s like onboarding through doing. The system doesn’t just give answers, it teaches staff where the information lives.”

This approach transforms how new employees learn. Instead of spending weeks digging through files or relying on oral tradition, they can immediately contribute to meaningful work, while simultaneously absorbing the organization’s history and culture.

Empowering Staff Efficiency: From Administrative Burden to Strategic Work

For most government employees, administrative work consumes far too much of the day. Dave Solaro from Washoe County knows this firsthand. “I’m an engineer,” he shared, “and I hire engineers to do engineering, not to spend half their day formatting reports.”

By developing an internal AI system with Madison AI, Washoe County automated many of its most repetitive administrative tasks. Staff can now generate staff reports, policy summaries, or board memos in minutes instead of hours. The AI writes in the county’s style and formatting standards, pulling in decades of historical data for context and accuracy.

This shift has saved Washoe County hundreds of staff hours each month. More importantly, it allows employees to focus on the critical thinking and decision-making that truly drive public value. Solaro emphasized, “The work of government matters, and we can make it easier, better, and faster for real.”

AI also helps eliminate silos. Many cities and counties store records in multiple systems. Building permits may be in one database, council minutes in another, and planning files somewhere else. Previously, staff needed to search each separately or rely on specialized knowledge to find the right data. Now, AI can traverse all these systems, synthesizing information across departments into one complete answer.

Supporting Retiring Public Servants: Capturing Experience Before It Walks Out the Door

Perhaps the most urgent challenge for local government is the wave of retirements already underway. Pirnejad pointed out that up to 40% of the government workforce is at or near retirement age. “These are our department heads, managers, and supervisors,” he stressed. “They carry decades of experience and judgment that’s impossible to replace overnight.”

AI offers a new way to capture that institutional wisdom. Instead of relying solely on written desk manuals or exit interviews, retiring staff can record short video or audio reflections on key decisions, processes, or lessons learned. Those transcripts can then be processed by AI, tagged, and added to the organization’s searchable knowledge base.

When done right, this process doesn’t just document procedures, it preserves the context behind them: why a decision was made, what trade-offs were considered, and how it aligned with community goals. That depth of understanding helps the next generation of public servants make informed choices while building on the past, not reinventing it.

Building Trust and Accuracy: The Importance of Using Your Own Data

Throughout the session, Olsen emphasized one critical principle: AI works best when it’s built on your data.

Generic tools like ChatGPT or open AI platforms pull from public internet sources, which may not be accurate or relevant to your jurisdiction. Custom, risk-averse AI models, sometimes referred to as “jurisdictions-in-a-box,” solve that problem by training only on verified public records such as meeting minutes, resolutions, and adopted policies.

This approach keeps data secure while ensuring the AI produces accurate, defensible results. Equally important, staff are trained to verify and interpret what the AI provides, reinforcing critical thinking rather than replacing it. “We don’t want employees to rely on it blindly,” said Solaro. “It’s a tool just like a calculator or Excel. You still need to know what the right answer looks like.”

Your Concrete Next Steps

Not sure how to approach your jurisdiction's AI journey? Here are some tips on how to move forward:

Start Small: Pilot with one department or function (i.e., clerk's office, HR, planning).
Data Foundation: Identify where key institutional knowledge lives (SharePoint, email, archives).
Privacy and Compliance: Use secure, government-grade Al (SOC2 / HIPAA / GCC).
Training and Adoption: Position Al as a trusted teammate, not a threat.
Iterate: Gather staff feedback to refine prompts and improve retrieval.

Erica Olsen said it best, “You can’t put the genie back in the bottle.” AI is here to stay, and for local governments willing to embrace it thoughtfully, it offers a once-in-a-generation opportunity to preserve institutional knowledge, enhance efficiency, and prepare the next era of public service leadership.

Watch ICMA TV to see more from the 2025 ICMA Annual Conference, and access the 2025 ICMA Annual Conference keynote sessions on the ICMA Learning Lab.