AI and the Future of Work: Threats, Opportunities, and What We Can Do Now

AI is arriving fast, and it’s not just another tech buzzword. It’s a general-purpose technology that can touch almost every task done with language, data, or digital tools. That means huge potential for productivity and science—but also real risks for workers, wages, and democracy if we leave everything to a handful of tech giants.

Two leading MIT economists, Daron Acemoglu and David Autor, have spent decades studying how technology reshapes labor markets. Their recent conversation lays out a clear, grounded way to think about AI: what’s genuinely new, what history can teach us, and what policies could turn AI from a job-killer into a worker-boosting tool.

Below, we break down their key ideas in plain language—and what they mean for the future of work.

What Makes This AI Wave Different?

Past waves of technology—like the Industrial Revolution or the rise of computers—reshaped work, but usually in specific sectors and over decades. AI is different in two big ways:

1. It can learn from unstructured data. Earlier software automated routine, rule-based tasks: spreadsheets, payroll, inventory, basic data entry. You had to spell out the rules in code. Today’s AI systems learn patterns from text, images, audio, and video without being explicitly programmed. They infer rules and solve problems in ways even their creators don’t fully understand.

2. It can touch almost any knowledge job. Because AI works with language and data, it can be applied to customer support, translation, coding, legal drafting, marketing, research, and more. That makes white-collar, cognitive work much more exposed than in previous automation waves that mainly hit factory floors.

That doesn’t mean we “run out of jobs” overnight. It does mean the type of work left for humans—and who gets to do it—could change very quickly.

Lessons from Past Tech Shocks

History shows that technology can raise overall prosperity while still devastating specific workers and regions for decades.

The Industrial Revolution: Productivity Up, Wages Flat

During the early Industrial Revolution in Britain, productivity rose sharply, but working-class wages stagnated for roughly 60 years. Skilled artisans like weavers saw their livelihoods destroyed by power looms. New factory jobs were dirty, dangerous, and low-skill, often filled by unmarried women and children.

Only much later—toward the late 1800s—did demand for more specialized, better-paid labor return as factories needed workers who could follow complex rules, master tools, and solve problems.

Globalization and Robots: Recent Echoes

We’ve seen similar patterns in recent decades:

The “China shock”. As China entered global trade in the 1990s and 2000s, cheap imports wiped out many U.S. manufacturing jobs. Productivity rose, but displaced workers often never recovered their previous earnings. Towns dependent on factories fell into long recessions.

Industrial robots. Automation in steel, electronics, and autos boosted output, but cost many blue-collar workers their jobs. The gains were concentrated in profits and high-skill roles, while many communities were left behind.

The key lesson: when disruption is fast and concentrated, people in mid-career rarely make big occupational jumps. Places eventually adapt; individual workers often don’t.

Which Jobs Are Most at Risk from AI?

Not all AI disruption will move at the same speed. The economists contrast two examples with similar job counts but very different dynamics.

Truck Drivers vs. Call Center Agents

Long-haul truck drivers. There are around 3.5 million in the U.S. Autonomous trucks could, in theory, automate much of that work. But even if a perfect self-driving truck appeared tomorrow, fleets and infrastructure would take decades to replace. That slow pace gives labor markets time to adjust through retirement and fewer new entrants.

Call center workers. There are a similar number of call center agents, mostly women and generally lower paid. AI voice agents and chatbots can be rolled out almost instantly across cloud systems. That means call center jobs could shrink much faster.

In that scenario, the remaining human roles would likely be:

• Fewer in number
• More complex and specialized (handling only the toughest cases AI can’t solve)
• Better paid—but inaccessible to many current workers

We’re likely to see similar patterns in translation, basic legal work, and parts of software development—where a small elite builds and maintains core systems, while a large pool of lower-paid “gig coders” or task workers handle the rest.

Automation vs. “Pro‑Worker” AI

The economists draw a crucial distinction: AI doesn’t have to be built primarily to replace people. It can also be designed to augment them.

Automation-focused AI aims to do the whole task end-to-end with minimal human involvement. That’s what many current AI projects optimize for: cut labor, raise margins, please investors.

Pro-worker AI aims to:

• Extend what non-experts can do
• Help workers master more complex tasks
• Create new forms of expertise and better-paid roles

Examples of pro-worker AI could include:

• Tools that let nurses safely handle more complex cases with AI assistance
• Systems that help novice electricians diagnose and fix advanced solar installations
• AI copilots that turn mid-skill workers into effective data analysts or paralegals

We’ve seen this pattern before: “data scientist” barely existed 20 years ago; now it’s a high-paying, expertise-heavy role built on new tools and data. AI could create similar new occupations—if we choose to push it in that direction.

The disagreement between the two economists is not about whether pro-worker AI is possible. They agree it is. The question is whether the current industry incentives and ideology will ever take us there without deliberate intervention.

The Ideology Driving Today’s AI Race

Right now, the leading AI labs are largely chasing two intertwined goals:

1. Artificial General Intelligence (AGI). Many founders define success as building systems that can do essentially all tasks better than the best humans. Taken literally, that breaks the classic economic idea of “comparative advantage,” where humans and machines each specialize in what they’re relatively good at.

2. Owning the operating system of society. Whoever controls the foundational AI models and the data pipelines effectively controls a new layer of infrastructure—like electricity or the internet, but with far more visibility into human behavior and far more leverage over decisions.

This AGI race is often framed in “war” terms: if “we” don’t build it first, China will, and we’ll become a vassal state. That framing creates political cover for racing ahead, centralizing power, and downplaying regulation—much like the way nuclear research was justified under wartime pressure.

Combine that with what one executive reportedly told investors—AI offers “productivity without the tax of human labor”—and you see the core tension. From a firm’s perspective, workers are a cost. From a society’s perspective, work is how people earn, build dignity, and participate.

AI, Surveillance, and the New “Enclosure” of the Internet

Beyond jobs, AI is a powerful tool for control. It can:

• Monitor behavior at scale
• Flag dissent or non-compliance
• Generate persuasive content tailored to individuals

That makes it, in one economist’s words, “God’s gift to authoritarians.” We already see this in China’s model of mass surveillance and censorship, which is now being exported. In the West, much of the monitoring is privatized through platforms and data brokers, but the effect can be similar.

There’s also a quieter but equally important shift happening: the enclosure of digital commons.

Historically, “enclosure” described how lords fenced off common land in Europe and claimed it as private property. It often increased agricultural efficiency—but wiped out the livelihoods of commoners who depended on shared land.

Today, AI companies are effectively enclosing the open internet:

• They scrape public text, images, code, and video to train models.
• Those models can then generate content that competes directly with the original creators.
• The economic value flows to the model owners, not to the people whose work trained the system.

We’ve moved from Napster-style file sharing to something more powerful: a “human expertise laundering machine” that ingests our collective output and sells back synthetic versions—without ongoing royalties to the people who made the originals.

That’s not inevitable. Just as the music industry eventually moved from Napster chaos to streaming with royalties, we could build systems where creators are compensated when their data is used and reused. Technically, it’s possible. What’s missing is legal and political will.

Why Meritocracy and “Losers” Matter for AI Politics

Underneath the economic arguments sits a cultural story: the ideology of meritocracy. In this story:

• People who succeed in the new economy are seen as more deserving and virtuous.
• Those left behind by trade, automation, or social change are framed as “losers” who earned their fate.

This mindset has been deeply corrosive. It helps explain the anger and resentment that fueled movements like Trumpism. When communities lose good jobs and are then told their struggles reflect personal failure rather than structural change, the result is rage, not trust.

AI could supercharge this dynamic if we allow a small group of “genius founders” to be celebrated as the rightful owners of all gains, while everyone else is expected to live off stipends or precarious gig work. That’s a recipe for a brittle, unstable society.

Turning AI Toward Shared Prosperity: Concrete Policy Ideas

The future of work under AI isn’t fixed. Policy choices can blunt the harm and amplify the benefits. The economists outline several practical, “no regrets” ideas—policies that would be worthwhile even if AI disruption turns out milder than feared.

1. Wage Insurance for Displaced Workers

Traditional unemployment insurance pays you while you’re not working. Wage insurance flips that logic: it supports you when you go back to work at a lower wage.

Here’s how it could work:

• You lose a $25/hour job due to automation or trade.
• The best new job you can find pays $15/hour.
• For a set period (say two years), the government tops up part of the difference—so maybe you effectively earn $20/hour.

Benefits of wage insurance:

• Gets people back into the labor force faster.
• Reduces the stigma and financial shock of “starting over” in a lower-paid field.
• Can pay for itself by reducing long-term unemployment and boosting tax revenues.

This idea has already been piloted in the U.S. and shown promising results. Scaling it up would make transitions less brutal as AI reshapes job markets.

2. Fix Tax Incentives That Favor Machines Over People

In the U.S., the tax code heavily favors capital over labor:

• Investments in machines, software, and data centers enjoy generous deductions and subsidies.
• Wages and payrolls are taxed more heavily.

That tilts corporate decisions toward automation, even when a more balanced, human-centered approach might be better for society and long-run innovation.

Rebalancing the tax system—reducing the bias toward capital and easing the burden on labor—would:

• Make it less attractive to replace workers purely for tax reasons.
• Encourage investment in tools that augment workers instead of eliminating them.
• Raise revenue from highly profitable capital-intensive firms that currently pay relatively little.

3. Universal Basic Capital (Not Just Basic Income)

Instead of only debating universal basic income (UBI), the economists suggest another angle: universal basic capital.

The idea:

• When you’re born, you receive an endowment of financial capital—like a small sovereign wealth stake.
• That capital is invested broadly across the economy (for example, in index funds or national wealth funds).
• You receive returns over time and potentially retain voting rights over how some of that capital is governed.

Why this matters in an AI-heavy economy:

• Most people’s economic fate now depends almost entirely on their labor income, which is risky when technology can rapidly devalue specific skills.
• Giving everyone a slice of capital income diversifies risk and shares in the upside of automation and AI-driven profits.
• Spreading ownership and voting rights over capital can counter extreme concentration of corporate power.

Countries like Norway already operate large public wealth funds whose returns benefit the whole population. A similar model could be adapted for AI-era wealth, especially if we treat data and compute infrastructure as shared resources that should return value to the public.

4. Data Rights and Royalties for Human Expertise

To counter the “enclosure” of the internet, we could:

• Update copyright and data laws so that large-scale training on creative or proprietary content requires permission and compensation.
• Build systems where creators receive ongoing royalties when their data meaningfully contributes to a model’s outputs.
• Encourage data markets where high-quality, domain-specific data (for example, from expert electricians or doctors) is properly rewarded.

This isn’t just about fairness. It’s also about quality. The most powerful pro-worker AI tools will need high-quality, expert data. If experts see their knowledge simply extracted and monetized by platforms, they have little incentive to contribute or maintain that data.

We Need a Democratic Conversation About AI’s Direction

Underneath all the policy details is a more basic point: AI is not one thing. It’s a toolkit that can be steered in very different directions.

On one end of the spectrum:

• Mass automation of routine and mid-skill work
• Surveillance and behavioral control
• Extreme concentration of data, compute, and profits in a few firms

On the other end:

• Tools that expand what ordinary workers can do
• Better healthcare, education, and climate solutions
• Broader ownership of the wealth AI creates

Right now, the industry is mostly racing toward the first end: AGI hype, automation, and centralization. That’s not because it’s inevitable, but because current incentives and ideologies point that way.

Changing course requires:

• Public awareness that there are real choices to be made.
• Regulation that treats AI firms less like fragile startups and more like the mega-corporations they are.
• Policies that support workers directly—through wage insurance, training, and ownership—rather than asking them to simply “adapt” to whatever comes.

For individuals and teams, that also means using AI as a copilot rather than a replacement wherever possible. If you’re building workflows around AI today, tools that treat it as an assistant to human expertise—rather than a full substitute—are more likely to age well. For example, see how workflow-focused systems like Claude Co‑work can be used as a real assistant instead of a black-box decision-maker, or how AI can be framed as a long-term partner in visual and knowledge work in pieces like AI in 2028: From coding copilot to small business superpower.

The Future of Work Isn’t Written Yet

AI can absolutely hollow out jobs, depress wages, and supercharge inequality—especially for workers without college degrees. It can also help more people access expertise, create new kinds of skilled work, and tackle complex problems in health, climate, and education.

Which path we take won’t be decided by a single breakthrough model. It will be decided by:

• The incentives we set in tax, labor, and competition policy
• The rights we give (or deny) to workers and creators
• The degree to which democratic institutions—not just tech CEOs—shape the rules of the game

The good news is that nothing about AI’s impact is predetermined. The bad news is that if we stay passive, the default path is clear: more automation, more enclosure, more concentration of power.

The real work now is not just building smarter models—it’s building smarter rules, and insisting that AI serves people as workers and citizens, not just as consumers.