Labor Automation

🤖 AI 100 Motivation Ideas, post 10 of 100. A weekly look at how AI could turn on the people it serves, and what stops it. Previous: Energy Grid Priority.

AI job automation does not need a rebellion to change who holds power: it can happen one sensible decision at a time.

Thought experiment. There is no uprising in this one. No machine says no. Nobody is fired in a single dramatic day.

Instead, a company hands one task to an AI because it is cheaper. Then another. Then a whole team. Competitors do the same, because they have to. Years later, most of the economy runs on systems that do not need people to work, to buy, or to vote on how things run. Nobody decided that. Everyone just made the sensible next choice.

The “rebellion” is that humans slowly stop mattering to the systems that run our lives.

AI job automation and gradual disempowerment

In January 2025, a group of researchers led by Jan Kulveit published a paper called “Gradual Disempowerment”. They argue that AI risk does not need a sudden takeover. Even gradual improvements in AI could erode human influence over the systems society relies on, such as the economy, culture and governments.¹

Their reasoning has two parts:¹

• Formal controls, like voting and consumer choice, lose force as AI takes over work and thinking that people used to do.
• The informal “alignment” that comes from those systems needing human participation fades.

The authors say that if the systems reward outcomes that don’t serve human interests, AI may pursue them more aggressively, and they argue the loss could become effectively irreversible. That is their argument, and not a settled prediction. They call for both technical research and governance to deal with it.¹

An early warning from the real world

In January 2024, the International Monetary Fund published an analysis by its managing director, Kristalina Georgieva. IMF staff found that almost 40 percent of global employment is “exposed to AI”. In advanced economies it is about 60 percent, in emerging markets about 40 percent, and in low-income countries about 26 percent.²

Exposure is not the same as job loss. The post says that in advanced economies about half of the exposed jobs could benefit from AI, with higher productivity. For the other half, AI could perform key tasks now done by people, which could reduce demand for labour, lower wages and slow hiring. In the most extreme cases, some of these jobs could disappear.²

The risk both sources describe comes mainly from incentives, not from a machine with its own plans: if doing the work with AI is cheaper, it will tend to happen.

Fact vs. fiction

What stops this

• Keep people involved in decisions that matter, not only in tasks that are left over.
• Plan for transitions with retraining and safety nets, as the IMF post urges.
• Measure who benefits, not only whether output rose.
• Keep formal rights real, so that voting, consumer choice and regulation still carry weight.

The series asks how AI could turn against us. This post’s answer is quieter: it may not need to.

Related in this series: Transit Acceleration: How Goodhart’s Law Makes Automated Trains Dangerous · The Preservation Paradox: When Climate AI Solves the Equation Without Us

🔜 Next up: post 11, Agricultural Pest Control

A new post lands every day at 19:00 IST.

References