Synthetic Labor: Agents and the Workforce of the Future
The language we use for AI lacks a crucial noun: labor. We need a more precise vocabulary.
When a paradigm-shifting technology emerges, our first instinct is to fit it into the world we already know. We reach for familiar metaphors, such as this is the “desktop” or the “cloud,” to make the novel understandable and accessible.
This tendency is particularly problematic when it comes to AI. For artificial intelligence, we have settled on terms that bring to mind ideas of assistance or delegation; we speak of AI “agents,” “copilots,” and “assistants.”
These terms are not wrong, but they are insufficient. They offer a useful, entry-level understanding of how humans interact with AI, but they lack the nuance required for deep strategic analysis.
By focusing exclusively on the technology’s function as a tool for a user, this vocabulary obscures its more profound economic identity (Chen, Srinivasan, & Zakerinia, 2024).
The language we use for AI lacks a crucial noun: labor. We are not merely building better tools but cultivating a new form of labor.
Synthetic Labor
“Synthetic Labor” as a Distinct Concept
Throughout this article, I use the term “synthetic labor”. It is the mental model I’ve been using; whenever I hear agent, API, copilot, or AI assistant, I’ve been substituting “artificial labor” and “synthetic labor”. I wanted to be very clear about the economic impact/role of integrating this new kind of labor.
While the academic and policy literature recognizes that AI and automation are transforming the nature of work and value creation, most sources refer to these phenomena using terms such as “automation,” “AI-driven labor,” “digital labor,” or simply “AI as a new factor of production” (Acemoglu & Restrepo, 2018; Tony Blair Institute, 2024).
These existing terms often emphasize the technological capabilities of AI or its impact on employment, but they do not fully capture the underlying economic identity of autonomous, non-human systems performing labor.
For example, the Harvard Business School working paper discusses how generative AI can displace or complement human labor, but does not explicitly frame AI as a new category of labor itself (Chen, Srinivasan, & Zakerinia, 2024). Similarly, the Tony Blair Institute report highlights the replacement of human labor by AI systems, yet it does not offer a specific term to describe the economic function these systems perform. By introducing the concept of synthetic labor, this article seeks to bridge this terminological gap. The most accurate and strategically useful term for what these systems produce is synthetic labor.
I define “Synthetic labor” as the autonomous execution of cognitive or physical tasks by non-human systems to generate economic value.
This term is intentionally chosen to distinguish the unique economic and strategic implications of AI-driven labor from both traditional human labor and the broader concept of automation.
Adopting this language enables a clearer analysis of how AI and robotics are reshaping economic structures, fiscal policy, and the distribution of wealth. It also prompts leaders to ask new questions about the measurement, management, and societal implications of this emerging form of labor.
Whether the “Agent” is augmenting, supplementing, or replacing labor done by employees, it is not just providing completed tasks or automation; rather, it is providing synthetic labor. The intelligence is artificial, but the labor is synthetic. When it is pure software, it is disembodied synthetic labor, and when artificial intelligence is embedded in robots, it is embodied synthetic labor.
As they integrate these solutions into their companies, companies are adding labor capacity, and in a very real sense, they are shifting the production of economic value in their organizations from biological carbon-based labor to synthetic silicon-based labor.
Synthetic labor, as opposed to human or biological labor, allows us to better understand the downstream consequences of synthetic labor at scale.
The Nuance Gap: From Augmentation to Autonomy
The term “agent” is insufficient because it conflates two fundamentally different economic functions: augmenting the work of a human and assuming the work of a human. A spreadsheet augments an accountant. The current generation of AI is capable of assuming the accountant’s core functions (Chen, Srinivasan, & Zakerinia, 2024).
This is not a theoretical distinction. Consider the fintech company Klarna, which in 2024 reported its AI system was handling a customer service workload equivalent to that of 700 full-time human employees. The system was not assisting 700 people; it was performing the economic function of their labor (Tony Blair Institute, 2024). This reveals the nuance gap in our current language.
To close this gap, we must differentiate the forms this new labor takes:
Cognitive Synthetic Labor: The invisible workforce delivered through software. It performs tasks like drafting legal briefs, analyzing market data, and writing software. Its key characteristic is unprecedented scalability, allowing for the execution of millions of cognitive task-hours at near-zero marginal cost (Holland & Davies, 2020).
Physical Synthetic Labor: The embodied workforce of advanced robotics. These are not the single-task arms of the past, but adaptive systems capable of navigating dynamic environments to perform intricate assembly, logistics, and maintenance tasks (Holland & Davies, 2020).
By classifying both as “synthetic labor,” we can analyze their collective economic impact with a clearer lens.
Three Economic Dynamics Our Language Obscures
Viewing AI as a new labor force reveals three profound, yet often understated, structural dynamics. These are not future crises to be feared, but present-day realities to be managed.
The Fiscal Dynamic: Decoupling Production from Public Revenue.
Modern fiscal architecture is built on a foundation of taxing wages. In the United States, for instance, individual income and payroll taxes provide roughly 85% of federal revenue, funding the core of the social contract. Synthetic labor operates largely outside this framework.
The value it generates flows not as taxable wages to an individual, but as revenue to the AI provider and efficiency gains for the adopter. This creates a quiet but persistent structural challenge: the economy can become vastly more productive while the tax base that supports society stagnates (Tony Blair Institute, 2024; Acemoglu & Restrepo, 2018).
The Demand Dynamic: The Paradox of the Non-Consuming Producer.
The 20th-century economy was powered by the virtuous cycle of wages enabling consumption. This symbiosis between mass production and mass purchasing power fueled growth. Synthetic labor is a supremely efficient producer, but it is not a consumer.
An AI model does not take out a mortgage, buy a car, or purchase services. As it shoulders a greater share of the economic workload, it systematically decouples value creation from broad-based purchasing power, posing a subtle, long-term challenge to aggregate demand (Tony Blair Institute, 2024; Acemoglu & Restrepo, 2018).
The Capital Dynamic: The Natural Path of Economic Returns.
The economic returns from human labor are, by their nature, distributed through salaries to millions of individuals. The returns from synthetic labor are, by their design, highly concentrated.
The value flows to the owners of the foundational models and the hyperscale infrastructure required to run them. With the top 10% of households already owning approximately 89% of U.S. stocks, this dynamic acts as a powerful engine for accelerating wealth concentration; not as a side effect, but as a direct consequence of the technology’s structure (Tony Blair Institute, 2024; Acemoglu & Restrepo, 2018).
The Strategic Imperative: From Insufficient Language to Expanded Vision
An insufficient vocabulary leads to an insufficient strategy. As long as we see AI primarily as an “agent,” our strategic responses will remain tactical and limited: we will focus on reskilling programs and productivity dashboards. These are necessary, but they do not address the systemic shifts underway (Tony Blair Institute, 2024).
Adopting the framework of synthetic labor elevates the conversation. It prompts a more sophisticated and essential set of questions for leadership:
How do we measure the full value and cost of deploying synthetic labor on our balance sheets, beyond immediate headcount reduction?
What is our organization’s role in the health of the consumer market upon which we depend, and how does mass automation affect that long-term dynamic?
If labor is no longer the primary factor of production and value distribution, what new mechanisms should we explore to ensure economic stability and broad prosperity?
The goal of language is to provide clarity, and clarity is the bedrock of sound strategy. The shift from “agent” to “synthetic labor” is more than a semantic debate; it is a mental model upgrade. It allows leaders to see the economic landscape as it is becoming, not simply as it has been, enabling them to move from a reactive posture to a proactive, intentional design
Works Cited
Acemoglu, D., & Restrepo, P. (2018). “Artificial Intelligence, Automation and Work.” NBER Working Paper №24196.
https://www.nber.org/system/files/working_papers/w24196/w24196.pdfChen, Wilbur Xinyuan, Srinivasan, Suraj, & Zakerinia, Saleh. (2024). “Displacement or Complementarity? The Labor Market Impact of Generative AI.” Harvard Business School Working Paper, №25–039.
https://www.hbs.edu/ris/download.aspx?name=25-039.pdfHolland, I., & Davies, J.A. (2020). “Automation in the life science research laboratory.” Frontiers in Bioengineering and Biotechnology.
https://www.pure.ed.ac.uk/ws/portalfiles/portal/177368167/AcceptedMS.pdfTony Blair Institute. (2024). “The Impact of AI on the Labour Market.”
https://institute.global/insights/economic-prosperity/the-impact-of-ai-on-the-labour-market