In an era where digital transformation is often synonymous with massive budgets and sprawling IT departments, small-scale government entities frequently find themselves caught between rising resident expectations and dwindling staff resources. Navigating this landscape requires more than just technical savvy; it demands a deep understanding of how policy, legislation, and operational discipline intersect with emerging technologies. As the leader of Government Curated, Donald Gainsborough has spent years at the helm of policy innovation, helping public sector leaders translate complex mandates into functional reality. His approach to artificial intelligence isn’t about chasing the latest shiny tool, but rather about hardening the operational foundations that allow a city or county to serve its people with integrity and speed.
The following discussion explores the strategic framework necessary for small governments to successfully adopt AI. We delve into why many initial pilots fail, the importance of categorizing implementation into distinct risk-based “lanes,” and the critical role of data hygiene. We also outline a pragmatic 90-day starter plan designed to move a team from informal usage to a disciplined, measurable pilot program that builds public trust.
When you observe small municipalities attempting to integrate artificial intelligence, what are the most common tactical errors that lead to these projects becoming “shelfware” rather than functional tools?
Small governments rarely fail because they lack ambition; they fail because they start their journey at the finish line. They often get seduced by a flashy vendor pitch or a high-level tool demo that promises to solve all their administrative woes, leading to a “pilot” program that eventually just sits on a digital shelf. This happens because the organization tries to implement a solution before they have a firm grip on their own internal processes, causing the team to conclude that AI is simply too risky or complicated for their already stretched-thin staff. We have to remember that AI is essentially a force multiplier for whatever current state your organization is in. If your internal processes are ad hoc and your data is a mess, the technology will simply automate that confusion and accelerate the chaos. The most successful entities are those that realize the fastest path to real value is not necessarily more software, but better, more disciplined operations that can be described, measured, and governed.
You advocate for a tripartite approach to AI implementation—separating productivity, process, and public-facing tools. Why is it so critical for a small district to sequence these specific “lanes” rather than tackling them all at once?
Sequencing is the difference between building a foundation and building a house of cards. We look at Lane 1 as staff productivity, which carries the lowest risk because it focuses on internal activities like summarizing long documents or organizing notes within approved, secure tools. This stage is vital for building “governance muscle memory” and reducing the natural fear staff might have about being replaced or making a catastrophic error. Lane 2 is where the real ROI lives, focusing on high-volume, rules-based process automation like accounts payable, procurement routing, or budget-to-actual reporting. By the time a government reaches Lane 3, which involves public-facing chatbots or decision-support systems, they have already established the necessary guardrails and audit logs to handle the high stakes of public trust. Jumping straight to Lane 3 without that experience is dangerous because small governments do not have the luxury of making a headline-grabbing mistake that could permanently erode resident confidence.
Before a single line of code is written or a subscription is purchased, you emphasize the “unsexy” work of process mapping. How does a clear understanding of current workflows prevent a government from experiencing “speed without steering”?
Without a clear map of your current workflow—including all those informal workarounds and “exceptions” that staff use to get things done—AI becomes a vehicle moving at high speed with no one at the wheel. Leadership must be able to answer exactly what a process looks like today and what specific data it touches before they can ever hope to automate it effectively. If you can’t describe the work or verify that the underlying data is accurate and protected, you are essentially asking the AI to guess, which is where errors and bias creep in. The irony is that doing this “unsexy” prerequisite work actually pays off even if you decide not to use AI immediately. It strengthens internal controls, reduces the need for rework, and makes it much easier to train new employees during periods of high turnover. It creates a baseline of operational excellence that allows any future technology to function as a tool of precision rather than a source of random output.
If a city manager is feeling the squeeze of staffing shortages and wants to see results quickly, what does a disciplined 90-day roadmap look like for a small team with limited resources?
A 90-day plan must be grounded in honesty and narrow focus to avoid overwhelming a small team. In the first 30 days, you have to get a real inventory of where AI is already being used informally and identify one or two major “pain” processes, such as permitting intake or invoice backlogs, to establish a baseline of cycle times and error rates. The second month, Days 31 to 60, is about putting guardrails in writing and picking a very narrow, rules-based pilot that requires human review, like an AI-assisted document routing system for accounts payable. Finally, in the last 30 days, you build and test using “golden test cases” that reflect the messy, real-world data we see every day, such as incomplete forms or ambiguous public records requests. This structured approach ensures that by the end of the three months, you aren’t just “trying” AI; you are running a live, monitored pilot with a clear rollback plan and measurable performance indicators.
Given the high stakes of public trust, which specific administrative functions offer the safest and most impactful starting point for a government’s first AI-assisted pilot?
I always point toward the “back office” functions—specifically finance and administration—because these areas are already governed by strict policies and produce outcomes that are easy to measure. Processing invoices, validating payroll adjustments, and tracking procurement compliance are ideal because they are high-volume, repetitive, and rules-based. These tasks touch data that we can easily classify and protect, and because they involve internal approvals, there is always a human in the loop to catch any discrepancies before they become public issues. By focusing on something like public records request intake and tracking, a small government can significantly reduce its administrative burden while maintaining a clear audit trail. This builds the necessary internal confidence to eventually move toward more complex use cases without risking legal exposure or the perception of unfairness in government decision-making.
What is your forecast for how small governments will evolve as they move past the “tool demo” phase and into mature AI governance?
I believe we are going to see a significant shift where the most successful small governments aren’t the ones with the most tools, but the ones with the best data discipline. In the coming years, “AI governance” will stop being viewed as a bureaucratic hurdle and start being recognized as a competitive advantage that allows municipalities to move faster and more safely than their peers. We will see a “flight to quality” where organizations treat their internal controls and transparency measures as accelerators for innovation rather than obstacles to it. Ultimately, the governments that thrive will be those that use automation to free up their staff for the judgment-heavy, high-empathy work that only humans can do, effectively turning their small size into an asset for agile, responsive, and deeply trusted public service.
