AI Agents and the Tax Base

Part of the Future of Work Series

Disclaimer: I create this content entirely on my own time, and the views expressed here are mine alone (not my employer’s). Because I love leveraging new tech, I use AI tools like Gemini, ChatGPT, Claude, Perplexity and others as a “digital team” to help research and polish these articles so I can share the best possible insights with you!

We talk a lot about what AI agents will do to jobs.

Will they replace workers? Will they make workers more productive? Will companies be able to operate with smaller teams? What happens when an AI agent can perform work that previously required three, ten, or perhaps hundreds of people?

Those are important questions.

But there is another question I don’t hear discussed nearly as much:

What happens to the tax base when the worker disappears?

And I don’t think this is only a question for Amazon, Microsoft, Walmart or some hypothetical Fortune 500 company eliminating 5,000 positions.

It may actually become more interesting when we look at small businesses and even one-person companies.

Because AI agents aren’t only giving large companies another way to reduce headcount.

They are giving small companies the ability to grow without adding headcount in the first place.

That could fundamentally change the relationship between economic growth, employment and government revenue.

The IMF Is Already Asking This Question

I originally started thinking about this as a fairly simple question about AI agents and taxes.

Then I came across something that made me realize the issue is much bigger.

The International Monetary Fund is already explicitly examining what it calls the erosion of the labor tax base as AI spreads through the economy.

In April 2026, the IMF published Global Economic and Financial Implications of Artificial Intelligence: Lessons from a Scenario Planning Exercise. The report examines different possible paths for AI adoption, including scenarios in which AI spreads rapidly through the economy and causes significant labor displacement.

One of the fiscal consequences identified by the IMF is remarkably close to the question I had been asking.

If AI reduces employment and wage income, governments collect less revenue from taxes tied to that labor.

The IMF puts it rather directly:

“Erosion of the labor tax base will necessitate adaptation of tax systems.”

That’s a pretty significant statement.

And the problem doesn’t necessarily go away simply because companies become more profitable.

AI can create enormous economic value while changing where that value appears, who receives it and where governments are able to tax it.

That last part is important.

Because there is a tempting counterargument to everything I’m describing:

If AI makes businesses more profitable, won’t governments simply collect more taxes from the businesses?

Maybe.

But it isn’t quite that simple.

Doesn’t Business Revenue Replace the Lost Tax Revenue?

Suppose AI allows a company to eliminate $400 million of payroll.

That doesn’t mean $400 million disappears. The economic value can move into several different places:

• AI infrastructure and cloud services
• Software and AI subscriptions
• Consultants and technology providers
• Additional corporate profit
• Business expansion and investment
• Higher compensation for some remaining employees
• Dividends or other returns to shareholders

Some of those activities generate taxes.

But business revenue is not the same thing as taxable profit.

A company can generate $10 million of additional revenue without generating $10 million of taxable income. Expenses still exist. Investments can be deducted or depreciated. Profits can be reinvested.

And the tax treatment depends heavily on how the business itself is structured.

That’s particularly important once we leave the world of large corporations.

Many small businesses don’t pay federal corporate income tax in the way we normally think about it. Sole proprietorships, partnerships, LLCs taxed as partnerships, and S corporations generally pass business income through to their owners, where it is taxed through the individual tax system.

So economic value hasn’t necessarily moved from “individual taxes” to “corporate taxes.”

It may have moved from wages paid to several people to business income received by one or two owners.

That’s a very different distribution of income.

And that brings me to small business.

This Isn’t Just a Fortune 500 Question

When we talk about AI replacing workers, our examples tend to be dramatic:

• A corporation eliminates 5,000 customer-service jobs.
• A bank automates a back-office operation.
• A technology company reduces its engineering staff.
• A retailer automates portions of customer service and administration.

Those examples are easy to visualize.

But small businesses are an enormous part of the American economy.

According to the U.S. Small Business Administration, small businesses account for nearly half of private-sector employment and more than a third of private-sector payroll.

So let’s make the example much smaller.

Imagine a marketing agency with 15 employees.

A few years ago, growing the company might have required hiring:

• Another graphic designer
• Another account manager
• Someone handling scheduling and administration
• Another copywriter
• Someone doing research and reporting

Now introduce AI agents.

The owner may discover that the company can grow from $2 million to $4 million in annual revenue while remaining at 15 employees.

Maybe it even operates with 12.

AI could handle portions of:

• First drafts
• Research
• Scheduling
• Customer follow-up
• Reporting
• Basic design variations
• Administrative work

From the owner’s perspective, that’s fantastic.

Revenue doubled. Profitability may have increased. The business became more productive.

And nobody necessarily got “replaced.”

But something economically important happened.

The jobs that would previously have been created never existed.

That may become one of the least visible effects of AI.

The Missing Hire

We tend to measure automation by layoffs.

Perhaps that’s the wrong measurement.

Consider a small accounting firm with eight employees.

Business increases 30%.

Historically, the owner might have hired two more people.

Instead, the firm deploys AI tools for things such as:

• Document intake
• Reconciliation
• Research
• Scheduling
• Client communications
• First-pass analysis

Revenue increases.

The company remains at eight employees.

There was no layoff announcement. Nobody filed for unemployment. Nothing appears in a headline about “AI replacing workers.”

But two potential jobs were never created.

Now multiply that decision across hundreds of thousands—or eventually millions—of small businesses.

That could matter just as much as the Fortune 500 company eliminating thousands of positions.

Then There Is the Solopreneur

This is where AI becomes even more interesting.

The United States already has tens of millions of businesses without paid employees. The Census Bureau calls them nonemployer businesses. Most are self-employed individuals, and sole proprietorships make up the large majority of them.

AI agents could dramatically expand what one person can do.

Imagine a consultant generating $150,000 per year.

Business grows.

Historically, that person might eventually have hired:

• An administrative assistant
• A salesperson
• A junior consultant
• Someone handling marketing or content
• Outside help for research and customer follow-up

Now imagine that instead the consultant uses AI agents:

• One manages scheduling and follow-ups.
• Another researches prospects.
• Another drafts proposals.
• Another maintains the CRM.
• Another produces marketing material.
• Another prepares first drafts of client deliverables.

The consultant might eventually build a $500,000 or even $1 million business with no employees at all.

That’s potentially an extraordinary productivity story.

I actually find that possibility exciting.

We’re giving individuals capabilities that once required an organization.

But from the perspective of the labor market and the tax base, something very interesting happened.

The business grew.

Revenue grew.

The owner’s income probably grew.

Taxes paid by the owner may grow substantially.

But the payroll never grew.

There are no employee payroll taxes associated with those potential jobs because those jobs were never created.

There are no additional workers receiving salaries and spending those salaries throughout their communities.

This is why simply asking whether business revenue replaces lost tax revenue misses part of the question.

One Million Dollars Is Still One Million Dollars — But Who Receives It Matters

Imagine two businesses that each generate $1 million of economic output.

Business A employs ten people.

Business B consists of one owner and a collection of AI agents.

Both might generate significant tax revenue.

But their economic footprints can be dramatically different.

Business A distributes income through ten paychecks. Those workers:

• Pay federal and potentially state income taxes
• Pay Social Security and Medicare taxes
• Receive wages across ten households
• Buy groceries, cars and other goods
• Pay mortgages or rent
• Purchase local services
• Spend money at restaurants and other businesses

The employer also contributes its share of applicable payroll taxes.

Business B may generate considerably more income for one owner.

That owner pays taxes too—perhaps quite a lot of them.

But the distribution is different.

There aren’t ten households receiving wages. There isn’t the same payroll-tax stream. There may not be the same volume or geographic distribution of consumer spending.

Some of the business’s expenses may instead flow to:

• Cloud providers
• AI companies
• Software platforms
• Data centers
• Technology consultants
• Infrastructure providers

Those companies may be located somewhere completely different.

The economic value didn’t disappear.

Its path through the economy changed.

And tax systems care enormously about where money flows.

Now Scale It Back Up

This brings us back to our large-company example. Imagine a company with 10,000 employees earning an average of $80,000. Over several years, AI agents allow it to operate with 5,000 employees. Five thousand jobs at $80,000 represents: $400 million in annual payroll.

Federal and state governments no longer receive the same taxes associated with those wages.

Social Security and Medicare no longer receive the same payroll contributions.

But those effects don’t stop at the worker.

The Worker Doesn’t Exist in an Economic Vacuum

Now remove thousands of workers.

The restaurant across the street doesn’t care that the company’s profit margin improved.

The coffee shop downstairs doesn’t receive part of the company’s AI productivity dividend.

The parking garage doesn’t sell parking spaces to AI agents.

And the office landlord can’t lease another floor to them.

This is where the tax effects begin spreading into the local economy.

And Then There Is the Office Building

If a company reduces its workforce substantially, someone eventually asks:

Why are we paying for all this office space?

Three buildings become two.

Two become one.

Entire floors may go unused.

That can affect:

• Commercial vacancy rates
• Lease revenue
• Commercial property values
• Property-tax assessments
• Building maintenance employment
• Security and cleaning contracts
• Nearby restaurants and retailers
• Public transportation ridership
• Parking revenue

A city doesn’t need to impose a local income tax to feel the effects.

The worker has an economic footprint extending far beyond the paycheck.

Remote work gave us an early glimpse of this.

With remote work, however, the employee often still existed. The paycheck still existed. Much of the spending simply moved from downtown to the neighborhood where the employee lived.

With AI displacement, the paycheck itself may disappear.

The Geography of AI Is Different

There’s another complication.

Imagine those workers were in Arizona.

Their salaries supported households and businesses in Arizona. Their economic activity generated tax revenue in Arizona.

Now imagine that much of their work is performed by AI inference running somewhere else:

• A data center in Texas
• A cloud region in Virginia
• Infrastructure elsewhere in the United States
• Potentially infrastructure outside the country

The company may still be located in Arizona.

Its customers may still be in Arizona.

But part of the economic infrastructure supporting the work has moved elsewhere.

This creates a fascinating problem for governments.

Economic value can increasingly be created somewhere other than where the company, customer or displaced worker is located.

Tax systems built around people, property and physical business locations weren’t necessarily designed for that.

Social Security and Medicare May Be the Bigger Question

This brings us to an even larger structural issue.

Social Security and Medicare depend heavily on payroll taxes.

Those systems were designed around an economy in which large numbers of people work and receive wages.

What happens if economic output continues increasing while payroll grows much more slowly?

That could happen through layoffs.

But it could also happen much more quietly through millions of missing hires:

• The 15-person company that never becomes a 25-person company
• The eight-person firm that grows 30% without adding two employees
• The five-person business that stays at five while doubling revenue
• The solopreneur who builds a million-dollar business without ever creating a payroll

None of these is necessarily bad.

In fact, from a productivity and entrepreneurship perspective, they could be tremendous successes.

But they raise a structural question:

If labor becomes a smaller percentage of economic production, should programs financed primarily through taxes on labor continue to depend on the same model?

So Should We Tax AI Agents?

This is probably where many people immediately go.

If an AI agent replaces a worker, perhaps the AI agent should somehow pay taxes.

Microsoft co-founder Bill Gates raised a similar idea years ago regarding robots.

I understand the logic.

But I’m not convinced that simply creating an “AI tax” is the answer.

For one thing, what exactly counts as an AI agent?

• Microsoft Copilot?
• A customer-service chatbot?
• A Python automation?
• An autonomous sales agent?
• A scheduling assistant?
• An AI system that makes one employee three times more productive without directly replacing anyone?

And what exactly would we tax?

• The model?
• The tokens?
• The GPU?
• The company using it?
• The company providing it?
• The amount of human labor theoretically replaced?

It becomes complicated very quickly.

The IMF has also cautioned against simplistic robot or AI taxes because poorly designed taxes could discourage productive investment while being extremely difficult to administer.

I think the more interesting policy question is broader.

Not: How do we tax a robot? but instead,

How should we tax an economy where an increasing share of productive capacity isn’t receiving a paycheck?

I suspect that is eventually where this discussion has to go.

This May Be the Bigger AI Transformation

I don’t want this to become another “AI is coming for your job” article.

AI agents could create enormous economic value.

They could allow:

• Companies to become substantially more productive
• Small businesses to compete with much larger organizations
• Entrepreneurs to build businesses with very small teams
• Workers to accomplish considerably more
• New products and services to emerge
• Entirely new industries and jobs to develop

That’s exciting.

But I’m increasingly convinced that focusing only on jobs replaced by AI may cause us to miss something bigger.

The future may not be dominated by companies firing half their employees overnight.

It may be much quieter:

• A company that once needed 100 people grows to the same size with 60.
• A 20-person business becomes a $10 million company without becoming a 50-person business.
• A five-person company stays at five.
• A solopreneur builds something that once required a department.

Nobody necessarily gets fired.

But fewer people get hired.

At the individual business level, that can be an enormous competitive advantage.

Across an entire economy, however, it could fundamentally change the relationship between productivity, employment, wages and government revenue.

Society’s Spreadsheet Has More Rows

When a business owner looks at an AI agent, the calculation can be relatively straightforward.

If I can accomplish more without hiring another employee, that’s good for my business.

I understand that calculation.

As a technology person, I’m fascinated by what AI agents could allow a small business—or even one individual—to accomplish.

But society’s spreadsheet has more rows.

An employee isn’t simply an expense.

That person is also:

• A taxpayer
• A consumer
• A commuter
• A homeowner or renter
• A customer of local businesses
• A contributor to Social Security and Medicare
• Part of the economic justification for offices, stores, transportation systems and other businesses

The answer isn’t to stop AI.

And I don’t think the answer is simply to “tax the robots.”

The question is more fundamental:

What happens when economic output and business revenue continue growing, but employment and payroll no longer grow with them?

We have spent decades treating economic growth, employment growth and tax revenue as closely connected.

AI agents could begin separating them.

A company can grow without hiring.

A small business can scale without building a large team.

A solopreneur can potentially create the output of what once required an entire company.

That’s an extraordinary opportunity.

But governments may eventually discover that a growing economy doesn’t necessarily produce tax revenue in the same places or in the same ways it once did.

The AI agent can do the work.

It can help generate the revenue.

It might allow one person to build a business that once required ten.

But it doesn’t get a paycheck.

It doesn’t stop for coffee on the way to work.

And that means the AI transformation may ultimately force us to rethink not only how we work—

but how we tax an economy in which work itself is changing.

What Do You Think?

I don’t pretend to have the answer to this.

AI agents are going to create tremendous opportunities for large companies, small businesses and especially solopreneurs. I think we’re only beginning to understand how much additional productivity they can unlock.

But if businesses can produce more while employing fewer people—or simply grow without making the hires they once would have made—the impact goes well beyond the business itself.

So I’m curious what others think.

• Does increased business revenue and profitability eventually replace the tax revenue associated with those missing jobs?
• If payroll becomes a smaller part of economic activity, how should we fund programs that depend heavily on payroll taxes?
• What happens to local businesses and communities when fewer workers are needed?
• Should governments change how they tax economic activity, or will the economy naturally adjust as new businesses and jobs emerge?
• And perhaps the biggest question: Are we paying enough attention to the jobs AI may prevent from being created, rather than only the jobs it replaces?

I don’t know exactly where this leads. But I think it’s a conversation we need to start having. What do you think?

References & Further Reading

• International Monetary Fund — Global Economic and Financial Implications of Artificial Intelligence: Lessons from a Scenario Planning Exercise
• International Monetary Fund — Broadening the Gains from Generative AI: The Role of Fiscal Policies
• U.S. Small Business Administration Office of Advocacy — Frequently Asked Questions About Small Business 2026
• U.S. Census Bureau — Nonemployer Business Characteristics
• Congressional Budget Office — Taxes
• U.S. Census Bureau — Government Taxes