AI Is Amplifying an Unexamined Government

AI—The Gamechanger

For decades, cities and counties operated on an understanding rarely discussed: municipal codes would never be fully enforced. I watched city councils adopt multi-family rental smoking bans and plastic utensil bans in response to zero public outcry, while staff confirmed there was no compelling local justification for the ordinance and no capacity to enforce it. I took those votes for grandstanding: a new power to punish residents who had harmed no one, adopted because adopting it felt good, or looked good, politically. The response was always the same. “Don’t worry. It will never be enforced, certainly not aggressively. Code enforcement operates on a complaint-only basis, and they can barely keep up with legitimate complaints.”

Then AI arrived and took the non-existent complainant’s place.

In Rancho Cordova, vehicle-mounted cameras now inspect 26,000 properties every month, flagging about 1,200 potential violations. Nothing about the municipal code changed. Nothing about the residents changed.

The limit was the enforcement budget the council adopted.

Councils passed ordinances knowing what that budget guaranteed. Rancho Cordova’s own code enforcement page sorts inspections by threat level, from an immediate danger to the public down to a violation with no health or safety hazard at all. That triage was the filter between the laws on the books and the laws residents actually lived under. The cameras flood the bottom tier, and the filter is gone.

Rancho Cordova is not alone. Cathedral City’s cameras flagged 623 violations across 36 violation types—vehicles on lawns, tarps on roofs, plywood over windows. Nobody has sorted the 36 into the ones that injure a neighbor and the ones that offend a preference of someone across town or working in city hall.

AI applications can be unquestionably good: patch the city’s code against attack, clear a permit backlog, answer the phone at 2 a.m. But a city buys capability, not judgment. Whether an activity should be performed was answered—or dodged—long before the purchase order, and nobody has to revisit that answer in order to do more of it.

AI Executes Whatever Mission it is Handed

A city’s costs are determined long before its budget is prepared. A mission states the organization’s purpose; the purpose sets the goals; the goals justify the activities. Most cities operate with a vague mission—“deliver services,” “maximize quality of life”—and a vague mission licenses goals without limit. Every activity those goals produce is either a cost to a resident or a business, or a use of the city’s power to compel: to tax, charge a fee, set a permit condition, issue a citation.

Cities buy AI at the end of that chain. The tools improve execution; the mission, the goals, and the list of activities all sit upstream, untouched.

AI expands what a city can afford to do. Every hour it saves becomes capacity for whatever was next in line—and what sits next in line is what could never justify a hire: the aesthetic code, the outreach initiative, the second compliance program. Nothing in that queue was ever sorted by whether the city should be doing it. AI just made the bottom of the queue affordable.

Productivity gains in a city do not leave the building. I watched a smaller version of the same conversion when city offices were automated: employees became more productive, and the payroll never shrank.

Code enforcement is only the most visible case. Every activity a city takes on was rationed by a cost of its own. When I worked for cities, hiring a grant writer was a real decision—a position, a salary line, and someone asking out loud which grants we ought to be chasing. The cost forced the conversation. Now the friction is trivial, and every grant won is a permanent program with matching obligations no resident voted to fund. Economic development prospecting, recreation marketing, and utility capital planning are all going the same way.

In June the State of California cut the price of AI tools in half for every city and county and built a portal to make buying them fast. “Reducing barriers,” the State Chief Information Officer called it. Expanding what a city does has never been cheaper, and nothing in the transaction asks which of those activities deserved the help.

Frozen Staffing, Expanding Scope

The same announcement that cut the price also froze the human side of the service delivery model. AI “should not replace the human work of government,” the Governor said; “it should help our workers move faster.” That guarantees the people and the speed but says nothing about scope. Los Angeles Unified School District’s agreement with its teachers union, approved in June, already makes the promise enforceable: generative AI may not replace bargaining unit work without the union’s written consent.

I am not arguing for a smaller payroll. A city with more employees confined to protecting rights is better than a city with fewer employees policing lawn height.

Hold the human element constant, add capability, and the gain has to surface somewhere. There are only three places it can go: back to the taxpayer, into how the work is done, or into more work. The first requires a council to decide the city is finished acquiring responsibilities, and I have never seen one decide that. The second is the variable everyone from the Governor to the bargaining table has promised not to touch. So productivity surfaces as scope. In code enforcement that means the code as enforced rising toward the code as written.

Elsewhere it means new work outright. A recreation software vendor asked 92 Arlington, Texas parks employees what they would do with hours given back: more programming, more community presence, more strategic planning. Nobody proposed returning the time. As the vendor put it, the constraint in recreation was never motivation—it was time.

The money runs the wrong way, too. Rancho Cordova is weighing raising the first-violation fine from $100 to $250, and up to $1,000 for repeat violations. Whatever the intent, one program now sets both the count and the price.

Cathedral City’s code enforcement director puts it plainly: “This isn’t going to take anybody’s job. It’s going to modify your workload.” Staffing fixed, workload changed—and the city is now inventing a job classification to absorb what the cameras produce.

A generation of municipal “reform” has meant changing how the work gets done without asking what the work is—contracting out, consolidation, regionalization, performance budgeting. All execution, and most with little to show by their own measures. No delivery model redeems an activity that shouldn’t be performed. But none of those reforms enlarged what a city enforces against residents who have harmed no one. This one does. How a service is delivered is a management question. What a city takes on is a question for the people who live under it, and it is moving now by purchase order, against no standard at all.

AI Passes Every Test That Doesn’t Matter

California now requires agencies adopting AI to examine civil rights and privacy, analyze impacts on marginalized and vulnerable communities, and plan for worker displacement. Rancho Cordova passes every one of them—uniform, discretion-free, displacing nobody. Equity review measures how a burden is distributed. It never asks whether the burden is legitimate.

A complaint-less citation has no injured party in it. The city is complainant, witness, and judge at its own discretion.

None of this defends selective enforcement; unevenness is a form of abuse. But the enforcement budget was doing the council’s work for it, and an ordinance a council would not enforce against every property should never have been adopted.

Mounting cameras that inspect every property in the city every month re-enacts every ordinance a camera can see. A second vote would not fix that; they were adopted by vote the first time. What no council ever applied to them is a standard: does enforcing this ordinance protect a resident, a business, or their property from a real harm, or does it impose one person’s preference on a neighbor who has harmed no one? Most cities are long overdue for a code review that asks that question of every local law—and then a second: would this council pass it knowing it will be enforced against every person, every property, every month, with no complainant?

Two questions belong in every AI staff report: which activity is this amplifying, and what right does it protect? “Code enforcement” is not an answer; “abating a structure that endangers the adjacent property” is.

For three decades the answer was always the same: “Don’t worry. This ordinance will never be enforced.” Today, the ones a camera can see are enforced against every property, every month. The rest are waiting on a detector that can see them.

Mark Moses is a senior fellow with the California Policy Center. He has thirty years of experience in local government administration and finance. His book, The Municipal Financial Crisis – A Framework for Understanding and Fixing Government Budgeting, was published by Palgrave Macmillan and is available from major online booksellers.

https://munifinanceguy.com/ · X/Twitter: @MuniFinanceGuy