Despite expectations that AI would be cheaper than human labor, this is not always the case today.

Training and operating advanced AI systems require massive investments in computing power, energy, and infrastructure. For many tasks, hiring a human remains more cost-effective than fully replacing them with AI.

The story of the Baxter robot illustrates this challenge. Although the technology appeared affordable, hidden costs such as maintenance, specialized support, and operational requirements made human workers a more flexible and economical choice.

Another overlooked factor is the cost of errors. Human mistakes are usually limited in scope, while AI errors can occur at scale, generating large amounts of incorrect or harmful output. As a result, the costs of supervision, risk management, and liability can significantly reduce the economic benefits of AI.

However, this advantage is likely temporary. As smaller AI models and more efficient hardware become available, the cost of digital labor is expected to decline substantially. Some analysts predict AI-related costs could fall by as much as 90% over the next decade.

Automation is expected to affect high-wage, aging economies first, including Japan, South Korea, and parts of Western Europe. In regions with lower labor costs, human workers are likely to remain competitive for much longer.

To prepare for this shift, societies should focus on developing skills that are difficult to automate, such as empathy, strategic thinking, creativity, and adaptability in complex environments. The key challenge is to use the current transition period to build an economy in which AI handles routine tasks while humans concentrate on uniquely human strengths.