Tax The Task

Every major technological revolution has required a corresponding revolution in institutional infrastructure. The industrial revolution gave us the payroll tax — because factory labor, unlike agrarian work, was concentrated, observable, and countable. The internet revolution gave us digital payment processing — because e-commerce, unlike cash transactions, needed new rails to move money across jurisdictions at the speed of a click. The advertising revolution gave us programmatic ad exchanges — because attention, once it became ambient and continuous rather than scheduled and discrete, needed new infrastructure to measure, price, and monetize it.

We are now entering the most consequential economic transformation since industrialization: the automation of physical labor by general-purpose robots and AI systems. And we have built exactly zero institutional infrastructure to govern it.

This is not a distant concern. In January 2025, Unitree demonstrated humanoid robots performing coordinated physical tasks with startling fluidity. By January 2026 — twelve months later — the same company's robots were virtually indistinguishable from human workers in controlled environments. Tesla's Optimus program is targeting a $30,000 price point for a general-purpose humanoid robot that can operate fifteen hours a day, every day of the year, at an effective cost of $1.46 per hour. The median American worker costs $27 per hour fully loaded. The math is not subtle. It is a six-to-one cost advantage that will reshape every labor-intensive industry on the planet within a decade.

The question is not whether this transformation will happen. It is whether we build the institutions to manage it before or after the fiscal and social consequences become unmanageable.

This paper argues that we need a new kind of infrastructure — a Task Clearinghouse — that makes robotic labor visible, classifiable, and governable. And it argues that the window to build it is closing fast.

The United States federal government collects approximately 80% of its revenue from sources directly tied to human wages: individual income taxes and payroll taxes. State governments depend on wage-linked revenue for 35-45% of their budgets. This is not a minor dependency. It is the load-bearing wall of modern government finance.

Every robot that replaces a human worker removes that worker's wages from the tax base. The employer stops paying FICA. The worker stops paying income tax. The state loses unemployment insurance contributions. The local economy loses the consumer spending that generated sales tax revenue. The effect is not hypothetical — it is arithmetic.

At a conservative estimate, if automation displaces 10% of the current U.S. wage base by 2035, the annual revenue gap is approximately $437 billion across all levels of government. That is larger than the entire budget of the Department of Veterans Affairs. At 20% displacement — plausible given the cost trajectories of humanoid robots — the gap exceeds $870 billion.

And here is the structural trap: the standard policy response to revenue shortfalls is to raise rates on the existing base. But you cannot raise payroll taxes on workers who no longer exist. You cannot collect income tax from wages that are no longer paid. The base itself is shrinking. Raising rates on a shrinking base is like running faster on a treadmill that's accelerating beneath you — the effort increases but the position doesn't change.

This is not a problem that resolves itself. Every year without institutional infrastructure for taxing automated production is a year the fiscal foundation erodes further. And unlike most policy challenges, this one has a ratchet effect: once firms automate and restructure around robotic labor, they do not reverse course when the government finally notices the revenue gap. The displacement is permanent. The fiscal damage accumulates.

The most commonly discussed response is a "robot tax" — a one-time levy on the purchase of robotic systems. Bill Gates proposed this in 2017, and the European Parliament debated and rejected it the same year. The robot tax fails for three reasons that illuminate why a fundamentally different approach is needed.

First, a one-time purchase tax captures none of the ongoing production value. A $30,000 robot operating for five years at fifteen hours per day executes millions of tasks. A one-time tax at the point of sale is like taxing a factory once when it's built and never again as it produces goods for decades. It is the wrong unit of taxation for a continuous production process.

Second, the revenue profile of a robot purchase tax is lumpy and procyclical — high during investment booms, low during recessions — precisely the opposite of what fiscal stability requires.

Wealth taxes, corporate profit taxes, and VATs have all been proposed as alternatives. Each has merit. None addresses the structural problem. Wealth taxes target existing fortunes but create no new revenue flow that grows with automation. Corporate taxes are constrained by international competition and can be restructured away. VATs tax consumption but are blind to whether the goods were produced by humans or machines.

The gap in the institutional landscape is specific: we have no mechanism to observe, classify, and tax automated production at the unit where it actually occurs — the task.

The key insight from two decades of labor economics research is that jobs are not the atomic unit of the economy. Tasks are. A hotel housekeeper's job is a bundle of tasks: stripping beds, vacuuming floors, cleaning bathrooms, restocking amenities, inspecting rooms, reporting maintenance issues. Automation does not replace the job in a single stroke. It replaces tasks, one at a time, until the remaining bundle is too thin to justify a full-time position. 

This is exactly how the smartphone consolidated consumer electronics. A 1991 RadioShack advertisement listed fifteen separate products — calculator, camcorder, tape recorder, CD player, personal computer, answering machine, mobile phone — each manufactured and sold as a standalone device. By 2012, every function on that page had been absorbed into a single object. The smartphone didn't improve the calculator. It absorbed the calculator's tasks into a general-purpose platform and eliminated the product category entirely.

Humanoid robots are doing the same thing to labor. They are not replacing one job with one machine. They are absorbing tasks across multiple job categories into a single general-purpose platform. And every task they absorb is a task that stops generating payroll tax revenue.

The task, therefore, is the right unit for fiscal infrastructure. Not the robot (too blunt, too hard to define). Not the job (too coarse, misses the gradual hollowing-out). Not the firm (too easy to restructure around). The task — the discrete unit of work that can be performed by either a human or a machine — is where the substitution actually happens, and it is where the fiscal system must meet it.

The Task Clearinghouse is a platform that sits between robot manufacturers, deploying firms, households, and governments. It performs 4 functions.

It classifies. Every automated task is categorized using a standardized taxonomy — the equivalent of the O*NET classification system for occupations, but for individual tasks performed by machines. "Vacuumed living room floor" is a different task from "folded laundry" is a different task from "sorted recycling." The taxonomy must be granular enough for fiscal precision and flexible enough to accommodate tasks that don't exist yet.

It counts. The platform tracks the volume of automated tasks performed across every connected system — not through surveillance, but through task metadata. Category, timestamp, duration, executing system. No video feeds. No audio. No personally identifiable information. Just the fiscal telemetry: what task, how long, which machine.

It reports. For commercial deployments (firms using robots in business operations), the platform generates compliance reports for tax authorities — automated filings that calculate the task tax owed, apply the correct rate for each task category, and remit payment. For the deploying firm, this is invisible. The platform handles it the way Stripe handles payment processing: the merchant doesn't think about interchange fees and PCI compliance. The infrastructure handles it.

It learns. This is where the architecture becomes powerful. The taxonomy is not static. It improves continuously through a reinforcement loop driven by millions of interactions across every connected system. And the most important source of training data is not commercial deployments. It is households.

The clearinghouse draws a sharp line between domestic and commercial robotic labor. The distinction is not incidental. It is the engine of the entire system.

Home use is free, untaxed, and privacy-protected. When a household robot vacuums your floor, folds your laundry, or loads your dishwasher, no tax is levied. The homeowner pays nothing to the clearinghouse. The robot manufacturer pays nothing. The domestic sphere is carved out entirely from the fiscal architecture — just as personal vehicle use is not subject to commercial vehicle taxes, and home cooking is not subject to restaurant health inspections.

But the domestic sphere is where the taxonomy gets built. The homeowner interacts with a dashboard — think Screen Time for your robot. "Your robot performed 847 tasks this week: 312 cleaning, 188 kitchen, 147 laundry, 200 maintenance." The homeowner can confirm, correct, or reclassify tasks. "That wasn't folding laundry — that was sorting recycling." Each correction trains the classification model. Each household is a free annotation engine, refining the taxonomy through natural interaction with a product they actually want to use.

This is the Google model. Google Search was free to consumers. Consumers never paid a cent. But every search query, every click, every correction trained the relevance algorithms that made Google's advertising product — the commercial layer — the most valuable business in history. The consumer was the training set. The advertiser was the customer.

The clearinghouse works identically. The homeowner is the training set. The government is the customer.

Business use is taxed through the clearinghouse at a percentage of collected revenue. When a hotel chain deploys 500 humanoid robots, each robot runs the clearinghouse SDK embedded in its operating system. The robots classify and report every task they perform. The clearinghouse calculates the task tax owed, files with the relevant jurisdiction, and remits payment. The government keeps 95% of collected revenue. The clearinghouse keeps 5% as the platform fee for building, operating, and maintaining the entire infrastructure.

The key insight is that by the time a robot manufacturer ships a unit to a Marriott or an Amazon warehouse, the taxonomy already knows what "clean a hotel room" looks like — because 50,000 households taught it. The commercial deployment doesn't need a custom classification project. It plugs into a pretrained system that has been refined by millions of domestic interactions. The domestic layer subsidizes the commercial layer's accuracy. The commercial layer funds the domestic layer's continued operation. The flywheel spins.

This architecture is not novel. It follows the proven playbook of every major platform business of the last fifteen years.

Airbnb did not build hotels. It built the infrastructure layer that made existing housing stock visible, bookable, and governable as short-term rentals. Before Airbnb, people rented spare rooms informally — an invisible, ungovernable market. Airbnb made it legible. It standardized listings, handled payments, collected and remitted occupancy taxes, and provided the data layer that allowed cities to regulate a market that previously operated in the shadows. The Task Clearinghouse does the same thing for robotic labor: it makes an invisible, ungovernable economic process legible to the institutions that need to see it.

Uber did not manufacture cars. It built the infrastructure that made private vehicle capacity visible, dispatchable, and taxable as commercial transportation. Before Uber, using a private car to transport strangers was illegal in most jurisdictions — not because regulators were hostile to the idea, but because there was no infrastructure to observe, price, and regulate it. Uber built that infrastructure and the regulatory framework evolved to accommodate it. The Task Clearinghouse follows the same logic: robotic labor is currently invisible to the fiscal system not because governments are indifferent to it, but because no infrastructure exists to observe and classify it. Build the infrastructure, and the regulatory framework will follow.

Stripe did not create commerce. It built the payment processing infrastructure that made internet transactions fiscally compliant by default. Before Stripe, every online merchant had to build its own payment integration, manage PCI compliance, handle multi-jurisdictional sales tax, and reconcile transactions across banking systems. Stripe made all of this invisible — a few lines of code and the infrastructure handled everything. The merchant didn't think about compliance. The compliance happened automatically at the platform layer. This is the exact model for the Task Clearinghouse: robot manufacturers embed the SDK, and the fiscal compliance happens automatically.

Google Analytics did not create websites. It built the measurement infrastructure that made website traffic visible, analyzable, and ultimately monetizable. Before Google Analytics, website operators had only crude server logs. Google gave them a free, powerful analytics dashboard that generated the data Google needed to build its advertising business. The domestic layer of the Task Clearinghouse follows this model precisely: give homeowners a free, useful robot analytics dashboard, and use the interaction data to build the taxonomy that powers the commercial layer.

The common thread across all four parallels is the same: in each case, a platform company built infrastructure that made a previously invisible economic process legible, and then captured value at the point of legibility. The Task Clearinghouse applies this logic to the largest invisible economic process of the 21st century: automated labor.

Robot costs are falling on a steep curve. A system priced at $30,000 today could plausibly fall below $10,000 within a decade, pushing effective hourly costs under $1. As the cost gap between human and machine labor widens, substitution becomes economically rational across more and more tasks. Even partial displacement of the wage base has massive fiscal implications.

If 40% of U.S. wages were displaced without new fiscal infrastructure, the annual revenue gap would approach $1.75 trillion. At 60%, it exceeds $2.6 trillion. The damage compounds annually, while the politics harden: once firms restructure around automated labor, retroactive taxation becomes far more difficult. The window to instrument the system is before it scales, not after.

The lesson from social media is clear. Platforms were allowed to scale globally before governance infrastructure existed. Retrofitting oversight proved slow, contentious, and incomplete. Robotic labor is at a similar inflection point. The companies are shipping now. Within a few years, millions of systems will be deployed across the economy. Fiscal architecture built after that moment will always be reactive.

The first principle is simple: infrastructure must precede entrenchment. The Task Clearinghouse makes robotic labor legible before it becomes invisible, and establishes fiscal alignment before crisis forces improvisation.

Even With An Upside Scenario

Even if automation expands the workforce rather than replaces it, the need for task-level accounting remains. If robots are additive, total output rises dramatically. But payroll taxation remains tied only to human wages, leaving machine production fiscally invisible.

In that world the problem is not collapse but misalignment. The economy grows, yet the tax base remains anchored to labor alone. Aligning public revenue with total production still requires visibility at the task layer.

The Task Clearinghouse addresses both futures: substitution or expansion.

The clearinghouse does not require new legislation to begin. It requires three things.

First, a pilot jurisdiction willing to experiment. A city or state that recognizes the fiscal exposure problem, sees the opportunity in being first, and has the institutional capacity to design a task tax framework in partnership with the platform. The pilot generates real data — actual task volumes, actual revenue, actual compliance rates — that becomes the proof of concept for every other jurisdiction.

Second, a relationship with at least one major robot manufacturer willing to embed the SDK. The manufacturer gets a differentiation feature ("robot analytics for homeowners"), a compliance solution for commercial deployments ("task tax handled automatically"), and early positioning as a responsible actor in a market that will inevitably face regulation. The manufacturer's incentive is not altruism. It is the same incentive that led Airbnb hosts to collect occupancy tax: compliance infrastructure makes the market legitimate and therefore larger.

Third, the taxonomy. This is the intellectual foundation — the equivalent of GAAP for financial accounting or XBRL for securities reporting. It must be open, extensible, and continuously refined by the domestic training loop described above. The taxonomy is the moat. Whoever defines the standard for classifying automated tasks owns the infrastructure layer for the entire robot economy.

The economics of the platform are straightforward. At a 5% take rate on task tax revenue collected through the clearinghouse, the platform reaches $1 billion in annual revenue when approximately $20 billion in task tax flows through the system. Based on the state-by-state fiscal model developed for this analysis, that threshold is crossed when roughly ten U.S. states adopt the task tax at moderate automation penetration — a scenario that is plausible by the early 2030s.

But the revenue model is secondary to the institutional function. The clearinghouse exists because the alternative — an unobserved, unclassified, untaxed robot economy growing exponentially while the fiscal foundations of government erode — is a calamity that compounds with every year of inaction. Building the infrastructure now, while the robot economy is still small enough to instrument from the ground up, is not just a business opportunity. It is an institutional necessity.

The industrial revolution gave us the payroll tax. The internet gave us payment processing. The attention economy gave us programmatic advertising. The robot economy needs the Task Clearinghouse. And it needs it before the window closes.