n.Newfeed

Business · 13 min read

How Platform Work Recreates the Company Town

A century ago, company towns controlled workers' homes, stores, and paychecks — then labor law supposedly ended all that. It didn't; it just moved to the cloud.

In the summer of 1894, George Pullman's employees discovered something clarifying about their lives. They had always known, in some abstract way, that they rented their homes from Pullman, bought their groceries from Pullman, attended a Pullman library, and received their wages in cash, with rent and company-store charges deducted from each paycheck. What the strike made viscerally clear was the logical conclusion of that arrangement: when Pullman cut wages by 25% without reducing rents, his workers had no floor to stand on. Their employer didn't just control their income. It controlled the entire material substrate of their existence.

The strike that followed became one of the most violent labor confrontations in American history. President Grover Cleveland sent federal troops. Eugene Debs went to prison. And the country drew a lesson that took decades to fully encode into law: no employer should be permitted to own the totality of a worker's life.

Now open your laptop and meet a Deel contractor in Manila, a Remote.com freelancer in Medellín, or an Upwork developer in Lagos. The geography has dissolved. The structure has not.


The Ghost of Pullman, Illinois Is Living in Your Laptop

The company town never died. It shed its geography, digitized its infrastructure, and slipped past every labor statute written to contain it. What replaced the physical town — the rows of brick houses, the company store with its marked-up flour and kerosene — is something simultaneously less visible and more total: an interlocking system of employer-controlled payment rails, curated benefit ecosystems, and algorithmic oversight that follows the worker not just to the factory gate but into their bathroom, their sleep schedule, and their search history.

This is not a metaphor deployed loosely for effect. It is a structural description. The legal scholar Sanjukta Paul has argued that the defining feature of the original company town was not its paternalism or even its coercion, but its completeness — the way it foreclosed alternatives at every layer of material life simultaneously. By that measure, the digital infrastructure now surrounding tens of millions of platform and gig workers achieves something Pullman's architects could only dream of: a company town with no zip code, no city council to petition, and no train you can board to somewhere else.

Understanding how we got here requires looking carefully at three interlocking mechanisms: the financialization of the paycheck itself, the consolidation of the benefit ecosystem, and the replacement of the human foreman with an algorithmic one. Each, taken alone, might look like a convenience or an innovation. Together, they form walls.


Scrip 2.0: When Your Paycheck Lives in a Corporate Wallet

George Pullman's scrip was elegant in its brutality. Workers received wages denominated in a currency that could only be spent in company-controlled stores. The conversion cost — the gap between scrip and real money — was invisible on payday and inescapable at the register. The modern equivalent is less legible but structurally identical.

Consider how payment actually flows to a contractor on a major platform today. Platforms like Deel route funds through proprietary wallets before they reach a worker's bank account, with withdrawal fees, currency conversion spreads, and processing delays that impose a steady tax on every transaction. Papaya Global and similar Employer of Record services charge workers fees for "expedited" access to money that is, legally speaking, already theirs. Coinbase Commerce integrations, adopted by several platforms as a cost-cutting measure on the employer side, push workers into crypto-denominated payments that require additional conversion steps — and additional fees — before they become groceries.

The more sophisticated version of this dynamic is the "Earned Wage Access" product, currently being marketed to HR departments across the country as a benefit. EWA programs allow workers to draw on wages they have already earned before the official payday — for a fee, sometimes expressed as a subscription, sometimes as a per-transaction charge, sometimes embedded so deeply in the product architecture that workers cannot identify it without reading a 40-page terms-of-service document. The Consumer Financial Protection Bureau's 2023 advisory flagged this explicitly, warning that EWA products operate in a regulatory gray zone affecting an estimated 7 million workers, and that no federal wage-and-hour statute was written anticipating a paycheck that is also a subscription product.

The company store's markup on flour was never more than 10 or 15 percent. The friction costs embedded in modern employer-controlled payment systems are harder to calculate precisely because they are deliberately distributed across multiple fee structures — but they function the same way. They make leaving expensive and staying subsidized, which is not a coincidence. It is a design principle.


The Four Walls of the Digital Benefit Ecosystem

If the payment system is the company store, the benefit ecosystem is the company physician, the company school, and the company church — rolled into a single dashboard and administered by people who also decide whether you keep your job.

Amazon Care, before its 2022 wind-down, was the purest expression of a trend that has not reversed: the employer-as-healthcare-provider. Walmart Health once operated clinics staffed by salaried employees serving a customer base that is, in many cases, also a workforce—until the company announced in April 2024 that it was closing them all. A growing roster of companies now offer vertically integrated health platforms where a worker's primary care physician, therapist, specialist referral network, and pharmacy benefits manager are all subsidiaries or contracted partners of the same entity signing their paycheck. This is a consolidation of dependency that exceeds anything Pullman's company physician achieved — he could only treat one illness at a time, in one town, with one set of records.

The mental health layer deserves particular attention. Employer-sponsored apps like Lyra Health and Spring Health have expanded rapidly, offering therapy and psychiatric care through platforms that are purchased by employers and deployed to workforces. The clinical data generated — diagnoses, session notes, behavioral assessments — is nominally protected by HIPAA-adjacent contractual frameworks. But "HIPAA-adjacent" is doing enormous work in that sentence. As health privacy scholars like Nicolas Terry have documented, behavioral and clinical data collected through employer-sponsored platforms exists in contractual gray zones where "de-identified" aggregates can legally inform workforce planning, performance benchmarking, and restructuring decisions. Workers often do not know this. They have been told, correctly, that their individual therapy notes are private. They have not been told what happens to the statistical shape of their anxiety.

The asymmetry of exit is where the modern company town most clearly reveals its architecture. Leaving Pullman in 1894 required a train ticket and a willingness to find work elsewhere — hard, but structurally straightforward. Leaving today's integrated benefit ecosystem mid-year means losing healthcare coverage for yourself and your dependents, potentially in the middle of an active treatment. It means repaying signing bonuses under clawback clauses that vest on employer-designed schedules. It means forfeiting unvested equity that may represent a significant fraction of your expected total compensation. Every one of these mechanisms was designed, individually, by a benefits consultant optimizing for retention. Together, they constitute financial gravity engineered to look like loyalty — and to be invisible to the labor statutes that were written assuming workers could, in principle, just leave.


The Algorithm as Foreman: Scoring, Surveillance, and the Invisible Supervisor

The factory foreman in 1920 was a figure of considerable power and considerable accountability. He could fire you, yes. He could also be fired. He had a name. You could appeal to his supervisor. Labor law, as it developed across the 20th century, was built around his existence — the assumption that disciplinary power in the workplace was exercised by an identifiable human being who could be made the subject of a grievance procedure, a wrongful termination suit, or a union negotiation.

The algorithmic foreman has none of these properties. Upwork's Job Success Score is a number generated by a proprietary model that the platform does not fully disclose, which directly controls whether a freelancer can access better-paying jobs or maintain their platform standing. Amazon's Time Off Task metric — the internal measure that tracks how long warehouse workers spend not scanning items — has been directly linked to termination decisions, with workers receiving automated warnings and, eventually, automated dismissals. Teleperformance, one of the world's largest business process outsourcing firms, deploys AI monitoring systems to its home-based agents that track facial expressions, background noise, and keyboard patterns in real time.

When a score drops and a contract is terminated under one of these systems, there is no manager to appeal to. There is no written notice required by law in most jurisdictions. There is no grievance procedure. The procedural protections that 20th-century labor law built around the moment of discipline — the right to know the reason, the right to respond, the right to have a representative present — were predicated on the assumption that discipline would be administered by a human being who could be held to those procedures. An algorithm is not a human being, and courts have been slow to find that it should be treated as one.

The behavioral reach of this surveillance is, if anything, the most striking finding in the empirical literature. A study by the Oxford Internet Institute found that 55% of platform workers reported modifying their personal behavior at home — eating habits, bathroom break timing, ambient noise levels in their living spaces — specifically to manage algorithmic perception. Workers were, in effect, performing compliance for an audience that existed only as a scoring function, reshaping their domestic lives around its inferred preferences. The drafters of the National Labor Relations Act, who were worried about foremen intimidating workers on the shop floor, did not imagine an oversight system that colonizes the worker's home without ever entering it.


Why the Regulators Can't See the Town They're Standing In

Labor law has a topology problem. The National Labor Relations Act of 1935—also known as the Wagner Act—and the Fair Labor Standards Act, the foundational architecture of American worker protection, were drafted around a specific picture of how work is organized: a single employer, a fixed workplace, a clock you punch, a supervisor you can name. That picture was never universally accurate, but it was accurate enough to anchor a legal framework that protected many millions of people for many decades.

The digital company town is deliberately topologically invisible. Its employer is often not, legally speaking, an employer at all — it is a "platform" connecting "independent contractors" to "clients." Its workplace has no address. Its clock is an algorithm. Its supervisor is a score. Every element of its structure was designed, or has been optimized over time, to fall outside the categories that labor law uses to impose obligations.

The "independent contractor" classification is the load-bearing wall of the entire edifice. It is not, in the strict sense, a loophole — loopholes are oversights. This is a deliberate legal architecture, maintained through sustained political effort. Uber, Lyft, DoorDash, and Instacart spent a combined $200 million defeating California's Proposition 22 in 2020, a ballot measure that would have reclassified their workers as employees. They spent it because the alternative was existential: reclassify the workers, and the legal exposure — for back wages, benefits, workers' compensation, unemployment insurance, the right to organize — becomes a number that threatens the business model itself. The lobbying is not disproportionate to the stakes. The stakes are genuinely that high.

The international dimension adds a layer that domestic regulatory reform cannot, by itself, address. Employer of Record services now allow companies to hire workers in 150 countries through intermediary entities headquartered in low-regulation jurisdictions — routing the employment relationship through wherever the legal obligations are lightest. A worker in Brazil, a worker in Nigeria, and a worker in Indonesia can all be employed, notionally, through the same Estonian shell structure, subject to the labor law of whichever jurisdiction offers the weakest protections for the transaction in question. This is regulatory arbitrage at planetary scale, and it is frequently announced to the market in press releases celebrating a company's commitment to "global talent" and "diverse, distributed teams." The DEI branding is not ironic, exactly. It is something more interesting: a genuine value, instrumentalized in service of a structure that systematically disadvantages the workers it claims to celebrate.


What a Modern Anti-Company-Town Framework Would Actually Require

Reform proposals in this space tend to come in two flavors: the incremental (extend FLSA coverage to gig workers, require fee disclosure on EWA products) and the visionary (universal basic income, platform cooperatives, algorithmic accountability boards). Both are necessary. Neither is sufficient without a clear diagnosis of what, exactly, the problem is.

The company town's defining harm was not that it paid workers badly, though it often did. It was that it engineered the foreclosure of alternatives — the systematic elimination of the worker's ability to exit, compare, accumulate leverage, or build a life independent of the employer's preferences. Any serious reform framework has to address that structural feature directly, rather than patching individual abuses while leaving the architecture intact.

Data portability is the most underrated intervention available. Workers should own their performance histories, reputation scores, professional credentials, and client feedback in standardized, portable formats they can carry between platforms — the same way a physician can carry their board certification, or a lawyer their bar admission, regardless of which hospital or firm employs them. An Upwork developer's five-star history should not be the property of Upwork. It is a record of that person's labor. Holding it hostage is the reputational equivalent of the company store: it makes leaving artificially expensive.

Algorithmic transparency requirements — mandating that platforms disclose the inputs, weights, and appeal procedures for any scoring system that affects a worker's access to income — would recreate, in digital form, the procedural protections that labor law built around human supervisors. It would not solve every problem. It would at least make the foreman visible again. And visibility, historically, is where accountability begins.

The benefit ecosystem requires a structural solution rather than a disclosure one. The consolidation of healthcare, mental health, and financial services under a single employer umbrella is not a problem that can be fixed by adding a consent checkbox. It requires either the decoupling of healthcare from employment entirely — the policy intervention that the United States has consistently refused — or binding fiduciary standards that prohibit employers from using benefit-generated data in any employment decision, with penalties severe enough to make the architecture unprofitable rather than merely discouraged.

None of this will happen quickly. The political economy of the digital company town is, if anything, more favorable to its continued existence than the political economy of the physical one was — because the workers it encompasses are distributed, often misclassified, frequently located in jurisdictions without the legal infrastructure to advocate for them, and separated from each other by the very platforms that employ them. George Pullman's workers could at least talk to each other on the walk to work.


Here is what the history of Pullman actually tells us, stripped of its nostalgic framing: the company town did not collapse because it was obviously unjust, or because workers successfully appealed to the conscience of the owning class. It collapsed because workers organized across the geographic barriers Pullman had tried to use against them, because a federal commission documented the structure of the dependency in embarrassing detail, and because enough political pressure accumulated to make the legal architecture that protected it more expensive to maintain than to reform. The digital company town is newer, faster, and better at making its walls look like benefits. But it is not, in any essential way, different. The question is not whether we can see it. The question is whether, having seen it, we will treat it with the same seriousness we eventually — after years of struggle, and considerable violence — treated Pullman.

Want the reader to remember this spot? Sign in — it is free.

Next What Four-Day Workweek Trials Measure and What They Leave Unresolved

The app

Keep going, or take it with you.

Free accounts read twenty pieces a day. Pro is US$1.99 a month or US$11.99 a year for unlimited reading.