What AI Makes Buildable

In the first essay, I argued that most AI writing mistakes the transition for the destination. The natural next question is what becomes possible on the other side.

By "buildable," I do not mean another toy app, another content mill, or a faster way to produce digital sludge. Humanity already has enough of that, and heaven knows the internet can survive without more.

I mean something more serious: organizations, services, and public systems that were previously too expensive, too brittle, or too administratively heavy to run well. AI's most important institutional effect, the one most commentary skips past, is that it makes coordination cheap.

For most of the modern era, scale came bundled with coordination. If you wanted reliability, scheduling, compliance, forecasting, and managerial visibility, you usually needed a larger organization. Big firms had many flaws, but they could afford the overhead. Smaller operators often lost not because they were worse at the underlying work, but because the back office quietly strangled them.

AI starts to unbundle that. When software can classify, draft, summarize, route, and monitor continuously at low marginal cost, some of the old advantages of size begin to leak outward. Scale no longer has a monopoly on coordination.

Thinner firms

A surprising amount of white-collar structure exists because organizations have trouble seeing themselves clearly. Information gets trapped in inboxes, shared drives, and tribal memory. Work has to be translated from one department into another. Managers spend too much of their time asking what is happening, who is waiting on what, and whether anyone has actually followed up.

A lot of bureaucracy is just an expensive way of helping the organization remember what it is doing. AI can shrink that burden considerably.

For decades, if you wanted serious operational sophistication, you needed layers: assistants, coordinators, analysts, middle managers, and the heroic few quietly holding everything together with duct tape and resentment. Now a 20-person firm can begin, in some respects, to behave more like a 100-person firm — not by replacing people with chatbots, but by collapsing the informational overhead that forced the headcount in the first place.

That matters for independence as much as efficiency. Firms can stay smaller longer, avoid premature bloat, and spend more energy on the work itself instead of the machinery required to manage it.

Stronger local operators

The clearest example may be the local service economy.

Take a plumbing company, an HVAC shop, an electrician, a roofer, a home-health provider. These businesses do not usually fail because they lack skill in the core work. They stall because everything around the work is chaos.

Calls are missed. Quotes are slow. Scheduling is messy. Crews arrive without the right materials. Customers do not know what is happening. Invoices go out late. The owner becomes a part-time dispatcher, part-time bookkeeper, part-time salesperson, and full-time bottleneck.

This is not a talent problem. It is a coordination problem. I see a version of this in my own work, building a product that replaces lockboxes with smart access for real estate showings. The old system is not stupid; at small volumes, it is workable. The friction shows up at scale. A buyer's agent wants to see a house. She calls or texts the listing agent. The listing agent checks with the seller. Someone reads out a lockbox code, often the same one shared with everyone showing the property. The showing happens, or it does not. No one is sure when. No one is sure who. If something goes wrong, the brokerage finds out late.

The technology to fix this is not exotic. The hard part was always the coordination layer: granting the right person access to the right place at the right time, with real accountability, without burying a small brokerage in administrative overhead they cannot afford. Smart access — the lock, the permissions, the logs — is software. AI becomes useful in the layer above it: parsing the buyer's agent's text in natural language, reconciling it against the seller's preferences, drafting status updates, and surfacing the cases that need a person. AI does not replace the listing agent's judgment. It removes the coordination drag around it.

That pattern repeats across much of the local trade and service economy. The national player has brand, buying power, and financing advantages. But the local firm's real disadvantage is bandwidth, and bandwidth is what gets cheaper when coordination costs fall.

Local operators carry something large systems lose: real accountability, local knowledge, and a stake in the community beyond quarterly metrics. If AI can close the administrative gap without forcing those operators to sell their independence to a platform, more communities get to keep competent local capacity. That is not a small thing.

Regional production and logistics

The same logic applies to the physical economy. A huge share of the real difficulty in manufacturing and logistics lies in quoting, procurement, demand planning, inventory, scheduling, quality documentation, and shipping coordination. The informational burden can be as punishing as the physical one.

That is part of why local and regional production so often struggle. Overseas labor and giant factories are cheaper, yes, but the administrative load of smaller-scale production is also brutal.

A small machine shop competes on speed as much as on cut quality. Today that competition starts the moment a quote request lands at four on a Tuesday. The material price moved last week. The foreman is on the floor. The next open machine slot is buried in someone's notebook. A shop where AI can read the drawing, flag the price drift, draft a quote, and propose a slot before a human reviews it starts to compete on responsiveness it could never afford to staff for. A regional food distributor that can forecast demand and optimize routes keeps more margin. A small manufacturer that can manage greater product complexity without drowning in overhead finds markets that were previously out of reach.

Not every product will come back. Not every supply chain will localize. Reality continues to contain atoms, trucks, weather, and human stubbornness. But the threshold at which local or regional production becomes economically sensible can move quite a lot once the coordination tax falls. Many of the twentieth century's geographic patterns were responses to the simple fact that coordinating many moving pieces was hard, slow, and expensive.

Public systems that actually function

Some of the highest-value uses of AI will not look futuristic. They will look like a permitting office that works, a licensing process that makes sense, a benefits system that does not force people to become their own case managers, a procurement office that can compare options without taking six months and three nervous breakdowns.

Public systems are often broken less by malice than by friction. Rules are scattered. Forms are opaque. Departments do not share context. Staff are overloaded. Cases come in incomplete. People do not know what is required, so they guess badly and pay in delay.

AI can help turn scattered rules into guided workflows, check applications for completeness before a human sees them, summarize case histories, surface relevant precedent, and translate administrative language into something a normal person can parse. Used properly, that is not a replacement for public judgment but a way of restoring public capacity.

The model should not quietly decide whether someone loses a business license or a disability benefit. Rights and obligations should not become a black box. But the tedious work surrounding those decisions — the assembly, the formatting, the routing, the status-tracking — can and should be reduced.

The state does not need to become all-knowing. It just needs to stop losing its own paperwork and punishing everyone for the privilege.

More human work, less administrative waste

There is a long list of professions where the real work is human and the administrative work around it has grown up to fill the space. Doctors, teachers, pastors, tradespeople, small business owners, lawyers — people whose real value lies in judgment, presence, trust, or skill.

That paperwork was not invented out of malice. It was invented because coordination used to be expensive. A hospital with a hundred caregivers and a thousand patients had no way to know who had been seen, what had been tried, or what was changing without an enormous apparatus of charting, scheduling, billing, and reporting. A school district could not see across its classrooms without standardized forms and reports. A small business could not stay legible to tax authorities or regulators without paperwork that turned every workflow into something documented, formatted, routed, archived. The administrative load was the price of operating at scale at all.

The math changes when intelligence becomes cheap. A doctor can have her notes drafted without losing half her morning to documentation. A teacher can stop losing hours to formatting and duplicative reporting. A small business can stay legible to the systems around it without its owner becoming a part-time bookkeeper. The administrative work was never the point. It was the toll people paid to operate at scale before there was a better way.

No serious person believes a chatbot is a substitute for a good doctor, a good teacher, or a steady tradesman. Sick people still need actual care. Children still need actual adults. The work that truly needs a person will still need a person. AI's most useful role here may be to retire the surrounding load — not to replace the work, but to let the work be the work again.

This does not happen automatically

Fine, perhaps AI makes these things possible. But why should we assume anyone will build them? Why would large firms not use the same tools to centralize more power, dominate smaller operators, and tighten control?

The objection is fair. The same coordination layer that helps a small firm stay lean can help a giant firm squeeze its suppliers harder. The same workflow tools that restore public capacity can multiply surveillance. None of this is automatic.

That is exactly why ownership and governance decide which version we get. Whether smaller actors keep real control, whether the systems are open enough to be adopted broadly or whether they tend toward concentration, whether the technology stays accessible at the household scale — these are not side issues. They are the issue.

Three questions worth asking

When someone claims an AI use case matters, I increasingly want to ask three unfashionable questions.

First: does it remove a real coordination failure, or does it just generate more output? Faster content is often faster clutter. More dashboards, more emails, more synthetic noise, and a slightly quicker way to annoy everyone is not a new operating model. It is spam with venture funding.

Second: does it increase the capability of smaller actors — small teams, local operators, public institutions — or does it deepen dependence on a central platform? A future in which the technology only strengthens the largest platforms is possible, but it is not the only one worth building toward.

Third: does it return human time to judgment, responsibility, and relationship, or does it just create a more efficient machine for monitoring people? If AI cuts five management layers but tightens the surveillance on whoever's left, that is not progress. It is leaner contempt.

The useful question

The useful question about an AI use case is what coordination failure it removes. If the answer is none, you are probably building more sludge. If the answer is real, the opportunity is more than a productivity gain — it may be a new operating model. It may change who can operate, who can compete, and what kinds of institutions become possible.

That is what AI makes buildable.

But productivity is not yet a future. Lower coordination costs do not by themselves produce a society anyone would want to live in. The next essay takes that up: what a hopeful vision of the AI age should say about work, dignity, ownership, and the shape of a good life once we stop treating people mainly as labor inputs.