By Futurist Thomas Frey

We are standing at the edge of a new industrial revolution. But unlike the last ones, which replaced human muscle or sped up information processing, this one is creating something fundamentally different: a global industry dedicated to manufacturing artificial workers.

These are not simple algorithms or reactive systems. They are AI agents—autonomous entities capable of perceiving, reasoning, planning, and acting with minimal human oversight. They don’t just assist us; they make independent decisions, generate value, and even create other agents. The question of the next decade isn’t just what AI will do, but who will build the systems that build the AI.

This is the paradox of the AI-agent economy: we are manufacturing the builders of our future workforce.

From Tools to Agents

Most AI tools today are reactive—they respond to inputs and deliver outputs. That’s powerful, but it isn’t agency. True AI agents can break down goals, devise strategies, and execute plans across multiple domains. Early experiments like AutoGPT showed glimpses of this. More recent agents from OpenAI, Anthropic, and Google have demonstrated the ability to research, write code, analyze markets, or run simulations with startling independence.

This is a leap from tools to workers. And industries are beginning to organize around that distinction.

Who Builds the Builders

The AI-agent economy is stratifying quickly:

• Foundation Model Giants: OpenAI, Anthropic, Google, and Meta provide the raw computational “brains.”
• Agent Architecture Firms: Startups like Adept, Fixie, and Cognition Labs transform these brains into decision-making frameworks.
• Domain-Specific Builders: Companies like Earth.ai (scientific discovery), ChatDev (software development), and Harvey (legal analysis) deploy agents into specific industries.

Already, Microsoft has embedded Copilot agents across its suite of productivity tools, while Salesforce’s Einstein GPT runs customer interactions without constant human supervision. Each case signals the same trend: the value layer isn’t just in models but in manufacturing the agents that apply them.

An Economy of Artificial Workers

Forecasts suggest the AI-agent economy could reach $1.5 trillion by 2030, with most of the value in enterprise productivity, consumer services, and specialized professional agents. That’s larger than the foundation model industry itself.

Unlike previous revolutions, the “product” here isn’t a tool—it’s a worker. The manufactured agents don’t just generate revenue; they expand the capacity of entire industries by multiplying the available cognitive labor force.

Work Transformed, Not Erased

Critics warn of mass unemployment, but history tells a different story. Mechanization didn’t end human labor—it redefined it. The AI-agent economy will do the same.

• Humans become orchestrators. Instead of executing tasks, workers will direct, audit, and refine AI teams.
• New jobs emerge. Roles like agent trainers, auditors, ethicists, and performance analysts are already in demand.
• Hybrid workflows take over. Teams of humans and agents will collaborate in ways neither could achieve alone.

The future of work won’t be about doing less—it will be about doing differently.

The Manufacturing Stack

Agent production is a layered process: foundation models, memory systems, planning frameworks, tool integration, safety guardrails, evaluation frameworks, and feedback loops. Each layer has its own specialists. Some companies own the entire vertical pipeline, while others innovate at a single critical point.

This stack isn’t just technical—it’s economic. The companies that dominate key layers will shape who gets to build, deploy, and profit from the next generation of artificial workers.

The Race to Control

The AI-agent economy is also a geopolitical contest. The U.S. leads in foundation models, China is advancing multimodal and embodied AI, and Europe is focusing on regulated domains like healthcare and finance. Export controls on chips and infrastructure reveal what’s at stake: agent manufacturing isn’t just about technology—it’s about sovereignty.

Meanwhile, decentralized platforms like HyperCycle are experimenting with democratizing access to agent manufacturing infrastructure. Instead of cloud monopolies, they’re building distributed networks that anyone can contribute to—or benefit from.

The Ownership Question

The deeper question is: who should own the means of production for artificial workers? Today, fewer than ten corporations control the most advanced capabilities. Should AI agents be treated as private assets, public utilities, or distributed commons?

The stakes couldn’t be higher. Whoever controls the builders controls the shape of future labor, economics, and power itself.

Final Thoughts

The AI-agent economy is not just another wave of automation. It’s the birth of a manufacturing industry where the output is intelligent, adaptive, and capable of building more of itself. This is manufacturing in its most recursive form: factories for artificial workers.

The question who builds the builders is more than philosophical. It is the central economic issue of the 21st century. The choices we make now—about concentration versus democratization, safety versus speed, orchestration versus exploitation—will define whether this revolution benefits the many or entrenches the few.

In building the builders, we are not just shaping the future of work. We are shaping the future of humanity’s relationship to intelligence itself.

Read more on related topics:

• Corporate “Reverse Acquihires” Are Reshaping AI Talent
• Edge Computing Is Creating a New Industrial Nervous System