AI may replace workers, create new demand

Artificial intelligence (AI) could replace workers in some tasks while increasing demand for labour in other parts of the production process, according to Canadian economist Jean-Louis Arcand.

The finding could have important implications for labour-intensive economies such as Bangladesh, where many workers are engaged in relatively low-productivity activities.

Tap here to add The Daily Star as a trusted source

Arcand, president of the Global Development Network, said firms are likely to adopt AI first in tasks that are the weakest links in a production process. By removing these bottlenecks, AI could improve productivity in other tasks and increase demand for workers.

“AI is potentially labour’s friend,” Arcand said at a public lecture organised by the South Asian Network on Economic Modeling (Sanem) at BRAC Centre yesterday.

The lecture, titled “These Aren’t the Droids You’re Looking For: Endogenous AI, O-Rings, and the Bottleneck Reallocation Theorem”, examined how firms decide where to use AI and how its adoption could affect workers.

Arcand’s research builds on economist Michael Kremer’s O-ring model of production, which says that production involves a series of interconnected tasks. A failure in one critical task can disrupt the entire production process.

He illustrated the idea with the 1986 Challenger space shuttle disaster, in which the failure of a relatively inexpensive O-ring contributed to the destruction of the spacecraft.

In his extended model, firms can use AI to reduce the risk of failure in specific tasks. This can affect employment in two different ways.

Within a particular task, AI can substitute for workers if it can perform the job more efficiently. Firms may then need fewer workers or less-skilled workers for that task.

At the same time, AI can complement workers in other tasks. When it removes a bottleneck, workers elsewhere in the production process can become more productive, increasing demand for their labour.

“When you introduce AI in a given task, it can crowd out labour, but it can increase demand for labour in other tasks because it is complementary to workers there,” he said.

Arcand said the impact of AI on employment would depend on the technology, the tasks involved and the cost of adoption.

“If someone asks you as an economist what will happen with AI in terms of labour, the correct answer is: it depends,” he said.

His model suggests that AI is likely to be adopted first in tasks where the risk of failure is relatively high. This means its use may not be concentrated only in advanced jobs or industries. Firms could first use it where it can remove major bottlenecks and improve productivity.

This could be relevant to service-sector activities such as coding and online customer service, where AI can reduce errors and improve performance.

The research also suggests that AI could help narrow wage differences. In the O-ring model, small differences in worker quality can lead to much larger differences in productivity and wages because the tasks are interconnected. Arcand said AI could reduce some of these differences by improving weaker tasks.

A calibration using US data estimated that the mechanism studied in the model could increase GDP by about 0.5 percent in the short term. However, Arcand said the gains were relatively modest, and the cost of adopting AI remained significant.

His analysis also suggests that lower-income economies could benefit more from AI than high-income economies.

However, countries such as Bangladesh will need to adopt AI strategically and use it to address genuine bottlenecks if they are to realise these potential gains, he said.

Arcand said more research was needed to understand the effects of AI adoption in developing economies. He also expressed interest in collecting data on AI use in Bangladesh.

“We can eventually collect data in Bangladesh on AI adoption,” he said.