Permitting lives in one system. Inspections in another. Finance in a third. GIS somewhere else. HR, code enforcement, public works, police, fire, utilities, planning, clerk’s offices—each with their own software, their own databases, their own filing conventions and often their own paper trails. Email, PDFs, scanned forms, handwritten notes and institutional memory held in the heads of people who are about to retire are still doing a remarkable amount of the work of keeping the system coherent.

This is not a modern, integrated digital environment. It is a patchwork.

And paradoxically, that makes local government one of the domains most structurally suited for AI.

Contemporary AI is not at its best when everything is clean, standardized and perfectly integrated. It is at its best when it can sit on top of messy systems: reading across formats, summarizing across silos, translating between schemas, reasoning over incomplete and inconsistent information and providing a semantic layer where technical integration has failed or never been attempted.

If any sector should benefit early and deeply from this kind of capability, it is cities.

So why hasn’t it?

Why AI Fits the City’s Technological Reality

Most municipal technology problems are not computational. They are connective.

Cities do not lack data. They lack coherence. They do not lack software. They lack interoperability. They do not lack procedures. They lack a way to see across them.

This is precisely where large language models and related systems are unusually well suited.

They can:

• Read ordinances, policies, council packets, staff reports and emails in natural language and answer questions that currently require knowing who to call.

• Act as a semantic layer across permitting, inspection and code enforcement systems that do not technically integrate but describe the same physical reality.

• Maintain institutional memory across turnover, capturing how things are actually done, rather than how they are documented.

• Route, triage and summarize resident requests before they ever reach a human, reducing the cognitive load on frontline staff.

• Translate between departments that use different vocabularies for the same problems.

• Surface patterns across 311 calls, police reports, planning cases and budget line items without requiring a unified data warehouse.

None of this requires a “smart city” rebuild. It does not require replacing core systems. It requires exactly what cities already are: heterogeneous, document-heavy, procedurally complex environments where meaning, not just data, is the bottleneck.

AI is not a futuristic overlay for cities. It is a compatibility layer for the present.

The Paradox of Low Adoption

Given this fit, the slow pace of meaningful AI deployment in local government is striking. Pilot chatbots exist. Some internal tools are being tested. Individual staff are quietly using general-purpose models to draft, summarize and research. But the kind of deep, structural augmentation one might expect has not materialized.

The reasons are not primarily technical.

They are institutional.

Political and Legal Friction

Local government is built around accountability, not optimization.

Every decision is reviewable. Every record is discoverable. Every mistake is litigable. Public trust is fragile and legitimacy is procedural as much as it is outcome-based.

An AI system that drafts a staff report is not just a productivity tool. It becomes part of the public record. An AI that answers a zoning question is no longer a convenience; it is a potential source of reliance and liability. An AI that routes complaints or prioritizes inspections is participating, however indirectly, in the exercise of state power.

This raises questions that are uncomfortable but unavoidable:

• Who is responsible when the system is wrong?

• How is bias audited when the output is probabilistic rather than rule-based?

• How is due process preserved when triage and summarization occur before human review?

• How do public records laws apply to model prompts, intermediate outputs and training data?

Cities are not allergic to technology. They are allergic to unbounded risk. And AI, particularly in its current form, feels like a new kind of risk: diffuse, opaque and difficult to explain in court.

Professional Norms and the Administrator Class

There is also a cultural barrier.

Professional administration, particularly in the council–manager model, has been shaped by a set of deeply internalized values: procedural legitimacy, incrementalism, risk aversion and continuity across political cycles. Organizations like ICMA have done an extraordinary job of institutionalizing these norms.

Administrators are trained not merely to manage, but to moderate. To slow change, to narrow the range of acceptable tools, to translate political ambition into legally and professionally defensible steps.

AI does not share that temperament.

By default, it accelerates. It collapses informational asymmetries. It reduces the cost of synthesis and comparison. It makes it easier for elected officials and department heads to ask “why not?” rather than “why now?”

This is not a technical disruption. It is a power shift. It weakens the traditional monopoly professional administrators have held over institutional memory, policy comparison and procedural navigation. It changes who can see across the system and how quickly they can act.

The Industry’s Mistake: Hype Over Glue

The technology sector has not helped.

For more than a decade, cities have been pitched visions of “smartness”: predictive policing, autonomous infrastructure, real-time dashboards, digital twins. These are impressive, expensive, and largely orthogonal to the day-to-day friction that actually consumes municipal capacity.

What cities need first is not more sensors or more platforms. They need connective tissue.

They need systems that can:

• Read their own documents.

• Understand their own procedures.

• Translate their own data.

• Preserve their own memory.

• Explain their own processes to residents and staff.

This is unglamorous work. It does not demo well in slide decks. It does not lend itself to ribbon cuttings. But it is exactly the kind of work modern AI excels at, and exactly the kind the industry has largely ignored in favor of selling futures instead of bridges.

Share

Why This Matters

The question is not whether AI can help cities. It clearly can.

The question is what kind of institution a city becomes when cognitive friction is reduced.

When synthesis is cheap, the balance between elected officials and professional staff shifts. When institutional memory is externalized, continuity no longer depends solely on tenure. When cross-departmental visibility increases, the slow, protective opacity of silos weakens.

This can be dangerous. It can also be liberating.

AI will not, by itself, make better policy. But it will change who is able to understand the system, how quickly they can do so and how much procedural mediation stands between intention and action.

That is why adoption is slow. Not because the tools are immature, but because the implications are structural.

Where Change Is Likely to Begin

The first deep transformations will not occur in the largest, best-resourced cities. They will occur where the pressures are greatest and the professional infrastructure is thinnest: small and mid-sized communities facing staffing shortages, retirement waves and rising service expectations with flat budgets.

There, AI will appear first not as “smart city” technology, but as:

• Institutional memory when people leave.

• Process interpreter when staff are stretched.

• Semantic glue across incompatible systems.

• Cognitive scaffolding for elected officials with limited administrative support.

Not as a replacement for human judgment, but as a layer that makes judgment possible at all.

Cities are not failing to adopt AI because they are technologically backward. They are hesitating because AI is not just another system to procure. It is a force that sits at the intersection of information, authority and legitimacy.

And that, more than any technical challenge, is what makes it transformative.

Subscribe now

Today, as AI systems and wearable technologies move from novelty to infrastructure, these old concerns feel newly relevant. But philosophy alone is not enough. We now have neuroscience, empirical insight into how effort, attention, and agency shape cognition. Taken together, these perspectives suggest that the real risk of AI is not that it replaces intelligence, but that it erases the signals our brains rely on to value effort, recognize agency, and engage authentically with the world.

Plato’s Fear, Reconsidered

Plato’s critique of writing, voiced through Socrates in Phaedrus, is often summarized too crudely as technophobia. His concern was not that writing was false, but that it altered the relationship between memory and knowledge. Writing stored information externally, allowing people to consult marks rather than cultivate recollection. Wisdom, he feared, would become detached from internal struggle.

History has not been kind to this worry in its literal form. Literacy is now one of the strongest protectors against cognitive decline. Reading demands inference, imagination, and memory reconstruction. Writing did not abolish cognition; it redistributed it. Plato’s mistake was not that he sensed a transformation, but that he lacked the biological tools to understand how brains adapt.

Still, his deeper intuition remains useful. Technologies that offload cognition change not just what we do, but how knowing feels. That phenomenological shift matters, even if the original fear was biologically misplaced.

Heidegger and the Disappearance of Effort

Heidegger pushes the question further. In Being and Time, he observes that tools disappear when they work. A hammer, when functioning properly, is not an object of attention but an extension of action, “ready-to-hand.” Only when it breaks does it reappear as a thing.

Wearable AI accelerates this disappearance. When perception, memory, and inference are quietly handled by a system that sees what we see and suggests what to think, cognition risks becoming invisible, not only to others, but to ourselves. The world is no longer encountered as something to be interpreted, but as something already pre-processed.

In Heidegger’s later work, technology does not merely assist; it “enframes,” revealing the world primarily as a resource to be optimized. AI systems that summarize, recommend, and decide can push us toward a similar mode of engagement: less interpretive, less dialogical, more consumptive.

Neuroscience and the Value of Effort

Where philosophy raises the question, neuroscience adds constraint. A growing body of evidence shows that effort matters, not morally, but biologically. Learning that involves retrieval, struggle, and prediction error produces stronger memory and deeper understanding than passive consumption. Cognitive reserve, the brain’s resilience against aging and decline, is built through sustained, effortful engagement.

Reading a book, even when it relies on external storage of words, forces active participation. Many AI interfaces do the opposite. They collapse process into outcome, removing the need to wrestle with uncertainty or construct meaning. When reasoning is outsourced wholesale, the neural systems responsible for it are exercised less, not more.

This does not mean AI is inherently harmful. It means that how AI is designed and used matters. Tools that reallocate effort, freeing us from drudgery while preserving interpretation, are very different from tools that remove effort entirely.

Authenticity as a Cognitive Signal

This is where authenticity enters the picture, not as a moral virtue, but as a cognitive signal.

People consistently prefer things that feel authentic: live music over perfect recordings, handwritten notes over flawless typography, art that shows its seams. These preferences are often dismissed as nostalgia or irrationality. But they make sense when viewed through the lens of effort attribution.

Authenticity bundles cues the brain is attuned to: evidence of constraint, traces of struggle, irreversibility, and human agency. These cues activate social-cognitive systems that deepen attention and engagement. They tell us that someone thought, chose, risked error.

In an economy saturated with polished, frictionless output, authenticity becomes scarce, and therefore valuable. It signals that effort occurred somewhere, even if we cannot see all of it. In this sense, authenticity functions as a proxy for meaningful engagement.

Wearables, AI, and the Risk of Erased Agency

The distinctive danger of advanced AI and wearables is not that they make thinking easier, but that they make effort invisible. When outputs arrive fully formed, instant, and indistinguishable from human labor, the brain loses its ability to differentiate between agency and automation. Once that distinction collapses, engagement shifts from relational to passive.

This is not a philosophical abstraction. If people cannot tell whether thinking happened—by whom, under what constraints, and at what cost—the incentive to think deeply erodes. Comfort replaces curiosity. Consumption replaces dialogue.

A Beacon, Not a Ban

The solution is not to reject AI or romanticize difficulty. Humanity has always adopted tools that extend cognition. The challenge is to preserve the signals that make cognition worth exercising.

Authenticity may serve as a kind of beacon here. Systems that preserve traces of effort, uncertainty, and human agency are more likely to support healthy cognition than those that erase them. The design question is not whether AI can think for us, but whether it can do so without obscuring the fact that thinking matters.

Plato worried that writing would hollow out wisdom. He was wrong about the outcome, but right about the stakes. Heidegger warned that technology could obscure our relationship to being. Neuroscience now shows that obscuring effort has measurable consequences.

The task ahead is not to slow intelligence, but to ensure that, even in an automated world, we can still recognize, and prefer, the marks of a mind at work.