Artificial intelligence is already finding its way into daily workflows for many employees, and necessitating others to think about the AI skills they’ll need to keep their jobs secure when companies embrace the technology.

AI may chip away at some of the most rote types of work, but experts have been clear AI won’t steal all our jobs. Most likely, they say, employees will work in concert with machines, and workers’ roles may become more sophisticated as AI eliminates repetitive and manual tasks. Ideally, employees will instead spend their time focusing on the uniquely human tasks that AI isn’t currently positioned to usurp – i.e. creative approaches to problem solving, interpersonal communication – and be able to generate more, better output.

If workers transition into more intellectually demanding roles, or see their productivity spike when they harness AI’s benefits, it’s tempting to believe wages will evolve in parallel. After all, if job descriptions are changing, and workers are contributing more, compensation should theoretically follow. 

However, experts say the outlook may be more complicated for workers of all skill levels. 

Going up? Going down? 

As AI takes some of the mundane tasks out of workers’ days, particularly in knowledge-work roles, experts expect many jobs may become more sophisticated and creative. Productivity could spike, too, as workers transition time-consuming minutiae to machines. For some employees, this means their fundamental job descriptions are changing – and that they’re producing more output at a higher level. 

Although it’s possible some companies will raise pay as employees take on new tasks, some experts, including Massachusetts Institute of Technology (MIT) economics professor Daron Acemoglu, are dubious all workers will see pay rises as AI pushes them to do more, different work.

In a 2019 research paper with colleague Pascual Restrepo, Acemoglu wrote that we “should not expect automation to create wage increases commensurate with productivity growth”. One reason, he believes, is that financial gains realised from increased worker productivity are generally absorbed by the company, not passed onto the workers as better compensation.

Harry J Holzer, economist and Georgetown University professor of public policy, and non-resident fellow at US think-tank The Brookings Institution, says this chimes with research that shows AI shifts compensation from the worker to the business. According to an economic study by the White House Council of Economic Advisors of automation in blue-collar jobs, Germany, France, Spain and The Netherlands saw revenue from productivity gains go to shareholders and not to the salaries of warehouse workers.

For hourly and shift-based workers, often in low-skill industries, AI has the potential to limit earning potential. Experts say one concern is workers may end up working less, as some parts of their jobs become automated.

Acemoglu beieves even though current automation technology may not match the quality of human work, companies will increasingly accept what he calls “so-so automation” in lieu of paying labourers – even if their wages are not substantial. He also warns this could become more of a problem as the technology gets increasingly advanced. 

And across all labour of all skill levels, increased worker monitoring may affect pay. Veena Dubal, professor of law at the University of California, Irvine, says many companies are already using AI to measure productivity, and subsequently determine wages.

Her research shows these evaluations lead to lower wages and inequality among worker earnings, which she calls “algorithmic wage discrimination – where individual workers make vastly different amounts based on an AI evaluation that the worker themselves doesn’t understand”.  She believes if companies invest heavily in productivity-tracking technologies, the problem could get worse.

There may be one group, however, that does stand to see their pay rise amid widespread AI integration.

Workers at top ranks, says Mark Muro, a senior fellow and policy director for the Metropolitan Policy Program at Brookings, may be more protected from wage fluctuation – especially if they facilitate and decide the processes by which AI is used in their workplaces. For instance, there may be “higher pay for some of those who are managing all of this [AI transformation] or doing the most sophisticated work supported by these tools”.

Levelling the playing field? 

The conversation around AI and salary isn’t just a question of whether wages will rise, stagnate or fall, however. Some experts argue that the rise of workplace AI could help close the wage inequality gap between high- and low-skilled workers, across many demographic groups.

David Autor, professor in economics at MIT, argues AI may be able to lower the barrier to entry for highly skilled positions by eliminating the need for a traditional, elite higher education usually required for a top role. With AI, a wide variety of workers can gain specialised expertise faster and more widely – in a sense, roles that were once gatekept may be increasingly less so.

“AI could create a scenario that reduces the scarcity of high-level expertise,” says Autor. As more workers from a variety of backgrounds work with AI to quickly gain the high-level skills usually reserved for specialised experts with elite backgrounds, “it would open the door for non-traditional candidates, particular those without a four-year degree”. 

He continues: “If the demand for highly specialised knowledge remains elastic” – for instance, in industries such as engineering, medicine, computer science and law – “that would mean more jobs for non-traditional candidates who, assisted by AI, would be able work at a high level of expertise. It could level the playing field.”

In this scenario, it’s possible the experts who were once scarce might see a dip in their wages as they see more competition in the labour market. However, it may be more likely that across industries, more workers – and a more diverse pool at that – would receive a higher salary than they might currently be making in lower skilled jobs that don’t require a college degree.

Organisation and advocation 

Muro says this may be a challenging period for every worker, as the technology is still developing.

MIT’s Acemoglu emphasises the outcomes for debates around worker pay and other AI-related workplace issues will largely depend on who is making the decisions about how the technology is developed and deployed, and what subsequent worker protections may follow.

Workers may have a part to play, says Dubal. 

To help drive these decisions in a human-favourable direction, she suggests workplace organisation and policy advocation may help workers keep compensation stable and protect their earning potential. For instance, if workers can “sit at the table with their employers and regulators”, they can help companies understand the intrinsic value of human labour. She points to European workers advocating “for laws that limit the use of AI in certain contexts, like the AI act that’s being developed”.

For now, it’s not clear which type of labourers will find their pay most affected by AI – after all, some experts have said highly skilled workers are just as vulnerable to automation as their lower-skilled counterparts. But, says Muro, “nobody gets off scot-free here”.

Additional reporting by Leah Carroll