Why We Need a New “Horsepower” for the AI Era: Introducing the Human Work (HW) Unit

In the late 18th century, Scottish engineer James Watt had a massive problem. He had just made significant improvements to the steam engine, creating a machine that was vastly more efficient than anything before it. But when he tried to sell his innovation to coal mine owners and farmers, he hit a wall of incomprehension. They didn’t care about cylinder pressure, joules, or thermodynamics. They only understood one metric of productivity: the draft horse.

To bridge this gap, Watt did something brilliant. He observed mill horses at work and calculated that a strong horse could lift 33,000 foot-pounds of weight per minute. He standardized this measurement and called it Horsepower (hp).

Watt didn’t invent horsepower as a pure scientific exercise; he invented it as a marketing and economic translation tool. It allowed a buyer to understand that one steam engine could instantly replace ten horses. That simple unit of measurement changed the global economy, giving society a way to quantify, value, and trade synthetic physical power.

Today, as we stand in the midst of the artificial intelligence and robotics revolution, we face the exact same translation problem. But this time, the machine isn’t replacing the horse. It is replacing us.

Until now, we have only created standard units to measure physical power (watts, horsepower) or computational speed (FLOPS, clock speed, tokens per second). But as AI transitions from a digital assistant to an autonomous worker, these metrics are completely failing to capture the economic reality of our new world.

It is time to introduce a new metric for the 21st-century economy: The Human Work (HW) Unit.

The Paradigm Shift: From Assistant to Replacer

To understand why the HW unit is necessary, we have to recognize how the nature of technology has fundamentally shifted.

For the last seventy years, the software and computing industries built “bicycles for the mind.” Word processors, spreadsheets, CRM platforms, and even early machine learning algorithms were all designed as amplifiers. They were tools wielded by human hands. Because humans were still doing the work, we continued to measure economic output in traditional terms: human labor hours and human productivity.

But modern Agentic AI and embodied humanoids are crossing a critical threshold. They are no longer productivity multipliers; they are outright replacers.

When you deploy an AI agent to research a competitor, draft a pricing strategy, and email a summary to your marketing team, the human is entirely removed from the execution loop. The AI is not assisting a junior analyst; it is the junior analyst. Recent 2026 studies into AI economics have already begun tracking “AI autonomy”—the degree to which users completely delegate decision-making and execution to models. We are witnessing the rise of the autonomous digital worker.

This reality breaks our traditional metrics. How do you measure the labor productivity of a marketing department if the software is doing 80% of the cognitive lifting? You cannot measure it in FLOPS (floating-point operations per second), because FLOPS measure machine exertion, not business value. And you cannot measure it in human hours, because the human hour has been eliminated.

The Humanities Roar Back: Why Tech Giants Are Hiring Liberal Arts Grads in the AI Era

Defining the Human Work (HW) Unit

Just as horsepower equated mechanical output to the labor of a biological horse, the Human Work (HW) unit equates synthetic cognitive or physical output to the labor of a biological human.

One HW unit is defined as the cognitive or physical labor an average, trained human professional can accomplish in one hour of focused work.

If an AI legal agent can ingest a 400-page contract, cross-reference it against state laws, and generate a flawlessly formatted risk-assessment brief in four minutes—a task that would typically take a mid-level corporate attorney eight hours—that AI has just generated 8 HWs of economic value.

The tech industry is already inching toward this conceptual framework. In July 2026, researchers evaluating complex software tasks proposed tracking AI progress using a “task completion time horizon”—a metric that explicitly measures AI capabilities against the time humans typically take to complete those exact same tasks. Anthropic’s recent Economic Index similarly tracks the “estimated human time to complete tasks without AI” to quantify the true impact of their models on the workforce.

By standardizing this concept into a universal HW unit, we create a quantifiable bridge between biological labor and synthetic labor. And as we will see, standardizing this measurement isn’t just an academic exercise—it is critical for the survival of our macro-economic systems.

Why the HW Unit is Essential for the Future

As AI systems and humanoids scale, capitalist economies are going to face unprecedented structural crises. The HW unit provides the mathematical foundation needed to solve three massive impending problems: GDP calculation, enterprise accounting, and taxation.

1. Rescuing GDP and Productivity Metrics

Labor productivity is traditionally calculated by dividing total economic output (GDP) by the total number of hours worked by humans.

Imagine an automated warehouse of the near future, managed by an AI logistics brain and operated entirely by humanoid robots. It runs 24/7, ships millions of dollars in goods, and employs zero humans. In this scenario, the “human labor hours” drop to zero. According to traditional economic formulas, the productivity of that warehouse approaches infinity—a mathematically useless conclusion.

If we don’t adapt, our national economic dashboards will go blind. By adopting the HW unit, economists can calculate the equivalent human work being performed by synthetic agents. Instead of tracking human hours, the Bureau of Labor Statistics could track HW output, allowing us to accurately measure the true labor engine driving the GDP, even when that labor has no heartbeat.

2. The AI Taxation Dilemma and Universal Basic Income (UBI)

This is arguably the most urgent reason the HW unit must be established. Modern capitalist governments are funded almost entirely by taxing biological humans. Income tax, payroll tax, Medicare, and Social Security contributions are all extracted directly from human wages.

What happens to a nation’s tax base when a massive corporation replaces 5,000 mid-level knowledge workers with a cluster of autonomous AI agents? The company’s profit margins skyrocket, but the government’s tax revenue collapses. The human workers lose their incomes, and the state loses the funds required to maintain roads, schools, and social safety nets.

Many technologists and economists argue that to survive this transition, society will need some form of Universal Basic Income (UBI) or a “robot tax.” But how do you tax an AI?

• You cannot easily tax “compute” or electricity usage, because that punishes companies for writing highly efficient code.
• You cannot easily tax software licenses, because a single AI agent can do the work of one person or one hundred people depending on how it is deployed.

The fairest and most accurate solution is to tax the value of the labor being replaced. If a corporation utilizes AI to generate 100,000 HWs of output in a quarter, the government can levy a micro-tax per HW. This ensures that as labor transitions from humans to machines, the economic abundance generated by those machines flows back into the societal infrastructure that supports the human population. The HW unit provides the exact taxable metric required to fund a post-labor transition.

3. Enterprise “Unit Economics” and Pricing

The software industry is already undergoing a massive pricing crisis. For two decades, software was sold on a “per-seat” SaaS (Software as a Service) model. You paid $50 per month for every human employee who needed access.

But what happens when AI agents eliminate the need for those human seats? If a customer service department shrinks from 200 humans to 10 humans overseeing an AI system, the software vendor’s revenue plummets.

To survive, the enterprise tech sector is rapidly pivoting to “AI Unit Economics.” Companies are abandoning per-seat pricing in favor of charging based on “Cost per Agent Run,” “Cost per resolved case,” or “Cost per completed task”.

However, measuring cost per agent run is chaotic because every vendor defines a “run” differently. By adopting the HW unit, enterprise procurement teams can establish a universal standard for ROI. A company will know exactly what it costs them to buy 1,000 HWs from OpenAI versus 1,000 HWs from Anthropic, allowing for transparent, competitive pricing in the synthetic labor market.

Conclusion: Driving the Future

When James Watt successfully marketed his steam engines using horsepower, he did more than just sell machinery. He gave society the psychological scaffolding needed to step out of the agricultural age and into the industrial age.

Today, you can walk into a dealership and buy a car with 400 horsepower without ever once picturing 400 actual horses pulling it down the highway. The metric transcended its biological origins and became a permanent fixture of our technological vocabulary.

Within the next fifty years, companies will routinely deploy AI servers capable of generating millions of Human Work (HW) units per second. Future generations will use the term without ever thinking about the biological humans of the 2020s whose cognitive labor set the original baseline.

But to get to that utopian future—a future where synthetic labor creates boundless economic abundance—we need a way to measure, tax, and distribute that labor today. The machines are already doing the work. It is time we start measuring it.

What are your thoughts on the Human Work (HW) unit? Could this be the key to implementing Universal Basic Income, or do you foresee challenges in standardizing the cognitive output of a human? Let us know in the comments below!