When I first started working in local government, I came in with a kind of idealism that only survives contact with a real institution for so long. I assumed government either worked or it didn’t: if a policy failed, it was because someone hadn’t tried hard enough, or hadn’t cared enough. What I found instead was a room full of people trying to solve enormously complex problems with limited budgets, competing interests, legal constraints, and imperfect information. Every decision I watched get made was a negotiation between what was ideal, what was legal, what was fundable, and what was politically survivable. Nobody in that room was lazy. Nobody was indifferent. They were constrained.

That experience didn’t make me trust government more blindly. If anything, it made me more skeptical of any process I couldn’t see. But it also reframed the question I was asking. I stopped asking “does government work?” and started asking “who gets to shape how it works, and when?” That shift — from judgment to participation — is the one I think we need to make now, at a much larger scale, as artificial intelligence moves from an abstract technology into physical infrastructure sitting in our neighborhoods.

People don’t lose trust in institutions all at once. It erodes in small increments: a meeting held at 2pm on a Tuesday when most residents work, a comment period nobody advertised, a decision that felt functionally made before the public hearing even began. Once trust erodes, cynicism sets in as a defense mechanism: if the outcome is predetermined, why bother showing up? I understand that reflex. I’ve felt it myself.

But here is the paradox I keep returning to: the answer to distrust isn’t disengagement. It’s better governance. Every time we opt out because we assume the process is broken, we hand the room over to whoever is willing to stay in it. Public meetings often feel inaccessible by design: the vocabulary of zoning ordinances, environmental review, and franchise agreements is dense, on purpose or by neglect, and it takes real effort to translate. It is far easier to criticize a decision after it’s made than to sit through the unglamorous work of shaping it beforehand. Participation is slow, often boring, and rarely rewarded socially the way outrage is. But it’s also the only mechanism we have for making sure institutions answer to the people who live with their outcomes. Disengagement doesn’t punish the institution. It just removes one more check on it.

This tension is playing out right now in Lancaster, where a proposed AI data center has become a live test of whether local government can actually govern something this large, this fast. Instead of asking whether the deal is simply good or bad, I think it’s more useful to look at what it addressed, and what it didn’t.

The agreement negotiated by the city took on several concrete concerns: water usage, given how much a facility like this can draw from a shared supply; noise, since these campuses run cooling systems around the clock; clean energy sourcing, since a facility this size can reshape the local grid; emergency response, because a large industrial site changes the risk profile of its surroundings; and community benefits, meant to make sure the neighborhood sees some return for hosting this kind of infrastructure.

That’s meaningfully more than many communities manage to negotiate. But the questions that matter most are the ones that come after signing. What happens after construction? Who measures compliance once the ribbon is cut? Who publishes the reports, and how often? Who responds if water use or noise levels exceed what was promised? Can residents actually see the data, or only summaries of it? These aren’t accusations. They’re design questions. And the answers will determine whether this agreement becomes a genuine template for how communities negotiate with AI infrastructure, or a one-time document that quietly stops mattering the moment the cameras leave.

Lancaster chose to negotiate. Not every community has. In Monterey Park, California, residents spent months building a campaign around a single ballot measure — Measure NDC, short for No Data Centers — after a developer proposed replacing an office building with a data center in their neighborhood. Volunteers went door to door with flyers in English, Spanish, and Chinese. The developer withdrew its application before the vote even happened. The city council banned data centers outright a few weeks after that. And when the measure itself went to voters, it passed with 86 percent in favor. The first citywide data center ban in the country decided at the ballot box. Monterey Park didn’t negotiate for better terms. It decided it didn’t want the deal at all, and it got there because enough people showed up early enough to have that conversation before a shovel ever hit the ground.

There’s a thought I keep coming back to, one that hit me almost offhand one day: people think AI is solving AI problems. I think AI is exposing human system problems.

AI isn’t creating trust — it’s revealing how little of it we had banked with our institutions to begin with. AI isn’t creating coordination — it’s exposing how uncoordinated our processes for infrastructure, land use, and utilities already were. AI isn’t creating civic participation — it’s exposing how thin our systems for gathering it have become. What AI is actually doing is forcing a question we’ve been able to avoid for a long time: can our institutions govern technology well, at the speed technology now moves? That question isn’t really about algorithms. It’s about zoning boards, utility commissions, procurement offices, and city councils — the unglamorous machinery of local governance — and whether that machinery was built for a moment like this one. In most places, it wasn’t. It was built for a slower century, with slower infrastructure and slower stakes. AI hasn’t broken that machinery. It has just made its age visible.

Here’s the part that stings a little: none of this is new information. Systems designers have been saying it. Community members have been saying it, often louder and longer. What was missing wasn’t the diagnosis. It was the willingness to actually run the numbers ourselves, in public, slowly, imperfectly, because that kind of math is uncomfortable, and discomfort is easy to avoid. (I’ll admit my own bias here: I’ve never had an easy relationship with math, even though it’s really just pattern recognition. The stakes just feel enormous when you’re the kind of person who wants every step to be right.) Now that there’s a giant calculator willing to do some of that math for us, everyone’s suddenly ready to move, mostly because that’s where the money is. What gets missed in the rush is that we already had a calculator for this. It’s called the collective.

There’s a phrase I use a lot in my work: the collective. It can sound abstract, even a little utopian, until you notice how much of daily life already runs on it. We invest collectively every single day, often without thinking of it as an ideology at all. Taxes. Roads. Public schools. Emergency services. Libraries. Retirement funds. Utilities. None of these exist because any one person built them. They exist because groups of people agreed, however imperfectly, to pool resources toward something none of them could build alone.

Maybe AI is reminding us that the biggest technology challenge in front of us was never intelligence. It’s coordination — and care. How do millions of people, with different needs and different levels of trust, coordinate and care for shared resources — land, water, power, attention? That has been humanity’s question for thousands of years, long before anyone imagined a data center. AI didn’t invent this problem. It just made it urgent again, and made the stakes harder to ignore.

I don’t think the answer here is simply “we need better government,” as a slogan. It’s more specific than that. Lancaster has an opportunity, not a guarantee. If this agreement becomes fully transparent, not just at signing but for the life of the facility. If residents stay involved after the initial hearings end. If outcomes are actually measured, and measured publicly. If companies are held accountable to what they promised, not just what they built. Then Lancaster could become something rarer than a cautionary tale: an example for how a community negotiates with AI infrastructure, rather than simply accepting it.

AI is not replacing the collective. It is revealing how much we depend on it.

Every question I’ve asked here — about water, about data, about trust, about participation — is really the same question asked six different ways: are we willing to show up for the systems we already own together? That was true before AI arrived. It will be true long after this particular data center is finished being new.