Researchers are warning of a gradual disempowerment
scenario where the widespread adoption of AI agents in the economy erodes human skills, creativity, and critical thinking. Unlike more vivid threats of rogue AI, this outcome stems from routine corporate incentives to automate labor, creating a society increasingly dependent on automated decision-making systems.
The Mechanics of Gradual Disempowerment
The concept of gradual disempowerment, which stems from a paper published in early 2025, describes a process where artificial intelligence displaces human labor and replaces decision-making in vital societal functions. As AI capabilities have surged since the 2022 launch of ChatGPT, businesses have increasingly turned to automated agents to handle tasks ranging from research and administrative work to specialized technical labor.
Economic Incentives at Hammer Co.
To illustrate the risk, researchers point to the hypothetical case of a manufacturing firm, Hammer Co.
If a company replaces its human staff with AI-powered robots, it can reduce costs and operate continuously without the need for breaks or sick leave. While a single business might prefer to retain its human workforce, competitive pressure forces a choice: adopt the automation or be outcompeted by rivals who do.
All the organizations have the incentives to adopt it and phase out humans. This is a very logical conclusion. It’s not very hard to see it.
Chooi
This dynamic extends beyond manufacturing to every sector, including state functions and cultural production. Because current economic and political systems incentivize efficiency and lower costs, the transition toward full automation is occurring without a deliberate, centralized plan.
Risks to Human Agency and Oversight
The danger, according to analysts, lies in the loss of human control over the systems that govern modern life.
Jan Kulveit, a co-author of the gradual disempowerment paper, highlights the political stakes of this shift. He argues that states historically maintain democratic structures because they require citizens for tax revenue and military service. If AI-driven systems render the human workforce unnecessary for economic and state functions, the incentive for governments to remain democratic may diminish.
Challenges in Measuring the Trend
Unlike threats such as cyberattacks or engineered pandemics, gradual disempowerment lacks the cinematic quality that typically captures public attention. Chooi notes that because there is no single, dramatic event, the trend remains largely outside the mainstream consciousness.
Proponents of this research argue that the first step in addressing the issue is to maintain indicators of how much work is being ceded to AI systems. Some argue that organizations must design systems that prioritize utility for people rather than simply aiming for high performance on technical benchmarks.
Whether society can reverse this path remains an open question. Because the transition is embedded in the complex interplay of culture, government, and the economy, proponents of the theory view it as a problem that will require a coordinated, whole-of-society approach to resolve.
