When discussing AI and labor markets, most people don’t think of IKEA. However, in a debate that often results in a lot of heat and comparatively little hard data, it might be the most helpful example. The company had to make a decision that many businesses in similar circumstances have not made when it automated a section of its customer support function with AI chatbots: instead of cutting staff, it reskilled the displaced employees as interior design consultants. According to IKEA, the outcome was a new source of income and a team of workers with truly more useful talents than they had previously. That was not an inevitable result. Deliberate investment, a clear resolve to not view automation savings as pure profit, and a readiness to bear the immediate expenses of retraining were all necessary.
Fundamentally, the question of whether AI can be pro-worker is about choices: who makes them, what motivations they have, and what limitations they are subject to. There is no preference in the labor market for the technology itself. Until someone determines whether to lower payroll by the same number of positions or redeploy the individuals who formerly performed that task, a technology that automates invoice processing is neither pro-worker nor anti-worker. Whether AI has a beneficial or negative overall impact on workers depends on the decision that is taken in that instant, repeated across thousands of businesses and millions of individual roles. Although it’s an honest remark, it may not sit well with those who prefer the solution to come from the technology itself.
Researchers and campaigners have begun building a pro-worker AI framework based on a few basic principles. Using AI to manage administrative overhead, routine analysis, and repetitive activities while leaving humans in charge of judgment, relationships, and complexity is known as augmentation over substitution, and it frequently results in better outcomes for employees as well as better outputs overall.
In the medical field, AI diagnostic assistance that maintains the doctor’s decision-making authority has demonstrated superior outcomes compared to systems intended to decrease human involvement. Coding tools that collaborate with engineers have continuously produced better reliable code in software development than entirely automated creation. According to the pattern, the human-in-the-loop model generates value that pure replacement does not, but it necessitates purposefully designing systems around it rather than just optimizing for cost savings.
One of the more prominent institutional initiatives to advance this strategy on a large scale is the RAISE US alliance, which has committed hundreds of millions to reskilling initiatives rather than merely acknowledging displacement as an economic externality. It is very unclear if that approach will extend beyond the businesses and sectors that have already made the commitment. Reskilling has a strong commercial rationale since skilled workers have institutional knowledge that is costly to replace, but in a cost-cutting climate, headcount reduction is more appealing.
Anyone should be wary of making firm forecasts in either direction because the larger economic argument is still unresolved. The optimists cite historical precedent, pointing out that previous technology revolutions eliminated certain types of job and created other ones, which were eventually absorbed by the economy.
The concerning aspect of pace and breadth is that AI is advancing more quickly than earlier disruptions, encroaching on white-collar jobs that were previously thought to be protected, and the transition time may result in significant expenses for individuals who lack the tools or access to handle it. The realists add a third observation: full automation of complex roles is farther off than the optimism suggests due to the practical limitations of current AI systems, including the ongoing issue of confident incorrect outputs, and enterprise adoption of new technology is almost always slower than headlines suggest.
In all three positions at the same time, there’s something worth clinging to. The result is not set in stone. There is the IKEA example. There are also cases of displacement. The issue of which becomes the norm is still up for debate and is unlikely to be resolved by technology alone.
