If you think your bots are going to somehow brute force City Hall, think again.


There's a line of thinking that governments will soon face a choice as AI agents overwhelm public services with requests. They can introduce new friction, making services harder to access, or use AI to increase their capacity.

We've been gutting public agency capacity for decades, because in the US we have trafficked the canard that civil servants are less capable, entitled and intentionally interested in following the rules.

To some degree, if your only interactions with government are being told no by someone unreasonable, you might agree. Hell, I've been previously been a public servant -- from City Hall to the Executive Branch -- and I would agree there are a ton of unnecessary blockers to getting useful things done, by people who seem to ignore reason and only want to cite to you the "law."

What bothers me about this framing is how little attention it pays to the status quo governments are actually operating under.

We've spent decades gutting public agency capacity, often justified by the idea that civil servants are incapable, unmotivated or perversely attached to following rules. I've worked in government from City Hall to the executive branch, and I know firsthand how many unnecessary obstacles stand between a reasonable request and a useful outcome.

In my Public Mechanics seminar, where we talk about the latency between policy and on-the-ground delivery, we discuss the difference between the actual law and what I call Bob's Law.

Bob's Law is the operating practice everyone treats as a legal requirement because of long-held office parables that might not actually be the rule, law or anything else. It's a personal preference passed down as lore.

Sometimes there's a law. Sometimes there's Bob.

Agents skipping the line are the Disney FastPass fantasy that people who believe lines in public agencies to be beneath them to skip the whole rigamarole of processes (that might exist for good reason) so they might have a few minute precious minutes in a day to diddle around with their neural devices or whatever.

The real problem with this line of thinking -- and I get that it's the prevailing wisdom of the industry right now -- that everyone needs to conform to the wishes of an entire sector freeloading off the rails of generations of public investment, is changing the ways we interact to accommodate fleeting technology, all to make it easier for people to cut in line would have catastrophic impacts on social & civic trust.

But there's another problem here. Mollick's formulation seems to imagine that governments can simply decide to offer faster processing for agent-generated requests. That would require an administrative and technical capacity many governments don't have.

Three-quarters of American incorporated municipalities have fewer than 5,000 residents. Many places are trying to maintain their existing services with limited staffing, aging systems and little money to replace them. A city can't create more inspectors or caseworkers by adding another way to submit a request.

And I struggle with the idea that governments ought to reorganize themselves around the demands of an emerging technology industry when we've already made it difficult for actual human beings to participate in their own government.

Office hours are a form of rationing. So are long appointment waits and backlogs that nobody has the resources to clear. Giving people with AI assistants another way around those constraints risks creating a Disney FastPass for public services. The people who can deploy agents get their own lane while everybody else continues waiting.

That requires investment in the people and systems that actually deliver public services. It means dealing with old procurement decisions and vendor contracts, understanding which rules are statutory and which are simply inherited practice, and giving agencies the technical staff to maintain what they build.

If AI companies want governments to support new kinds of automated interactions, there's plenty of work to be done inside those agencies. Forward-deployed teams that work with public servants to build lasting capabilities would be a more serious commitment than telling governments they need to accommodate another wave of requests.

Public agencies have obligations to their residents, those obligations don't change because someone has invented a faster way to ask.