5 flaws in the ‘augmented work’ myth

We’re being promised a revolution. Artificial intelligence is said to make work both more productive and more enjoyable by freeing us from tedious tasks and allowing us to devote more time to what really matters. This is the narrative of “augmented work.”

But what is the reality for people who are already working with AI?

To find out, our research team, led by the International Observatory on the Social Impacts of AI and Digital Technology (Obvia), partnered with 11 labour unions representing more than 1.4 million members. Together, we surveyed 4,595 union members in Québec in 2025. Our findings urge us to seriously reconsider the myth that machines universally improve work.

PRODUCTIVITY GAINS NOT AS ADVERTISED

Let’s start with the flagship promise: productivity. If AI lives up to its promises, its users should be able to work more efficiently. However, among those who use it daily, only 44 per cent report a real gain in productivity. Even more surprisingly, 26 per cent report a loss of efficiency because of the technology.

These productivity losses stem in particular from the proliferation of “quick and sloppy” work generated by AI (known as “AI slop”), whose poor quality leads to sometimes significant time losses. This could involve a report that needs to be redone because it does not meet quality standards, information that needs to be verified because it is deemed unreliable, a chain of emails that has become incomprehensible and requires clarification among colleagues, and so on.

Adopting AI therefore doesn’t guarantee real benefits. It all depends on the tasks, the context and the person’s profile. Moreover, AI remains, for now, limited to minor tasks: 75 per cent of users employ it for support or in a secondary role, and only three per cent make it a central part of their work.

AS MANY WINNERS AS LOSERS

The second promise of “augmented work” is a lighter workload. Here again, the picture is mixed. AI does indeed lighten the burden for 46 per cent of respondents. But it increases the burden for nearly one in five (17 per cent) and makes no difference for the remaining third. For example, a paralegal may become more efficient thanks to AI, but ultimately be assigned more cases or tasks, resulting in a heavier workload.

Our data shows the same ambivalence regarding workplace stress. AI reduces stress for 28 per cent of respondents and increases it for another 28 per cent, an even split. A college professor sums up this tension:

“Tools are evolving so quickly that we don’t have time to keep up with the changes. This is anxiety-provoking and discouraging.”

In other words, AI doesn’t improve work across the board, it reshapes it. And depending on the sector, profession or age of the employee, this reshaping can be for better or for worse.

IT DEPENDS ON WHO YOU ASK

This is undoubtedly the most important finding of our study: the benefits of AI are not distributed randomly; they follow the existing fault lines in the labour market.

Professionals and college graduates are among the biggest winners. They report using AI independently, becoming more productive and participating fully in decision-making. At the other end of the spectrum, technical, industrial and service workers are more often forced to use AI. They report experiencing greater algorithmic surveillance and seeing fewer of its benefits.

The perception of increased productivity clearly illustrates this divide: 55 per cent of post-graduate degree holders consider themselves more productive thanks to AI, compared to only 22 per cent of high school graduates.

LACK OF OVERSIGHT AND COMMUNICATION

There are two things almost everyone laments.

First, a lack of oversight within organizations: only 24 per cent of respondents say their organization has a policy or other mechanism governing AI use.

Second, the rollout is anything but transparent and democratic: barely 12 per cent say they were consulted before deployment, and only 26 per cent believe their organization is transparent about its use of AI.

As a result, only 35 per cent trust their employer on its AI initiatives.

FEAR OF JOB REPLACEMENT

Nearly one in five people (18 per cent) fear for their job because of AI. This concern increases with age and is significantly more pronounced among those working in manufacturing and services, and among those with fewer educational credentials.

These results contrast with most major forecasting studies that predict that skilled office jobs are the most at risk from AI.

This discrepancy between the models’ predictions and the fears people actually feel is worth examining: it serves as a reminder that technological anxiety often takes root where people already feel most vulnerable.

LOOKING AT THE FACTS

What should we take away from this? That “augmented work” is not a lie, but an incomplete narrative.

For some workers, often the most skilled, AI is indeed a welcome support tool that they use on their own initiative. For many others, it is a source of pressure, reduced autonomy, surveillance or job insecurity.

At a time when organizations are facing challenges related to skills, qualifications, staff shortages and innovation, the issues raised by AI in the workplace must be put back at the centre of decision-making: AI alone cannot be the solution.

Taking the time to discuss AI with all affected staff can lead to much greater results: the more people are trained, engaged and truly involved, the more tangible the benefits become, and the more we can prevent AI from ultimately undermining the very work it claims to improve in the absence of collective thinking about how it’s used.

» Xavier Parent-Rocheleau is a professor at HEC Montréal. Julie Garneau is a professor of industrial relations, Université du Québec en Outaouais (UQO). Vincent Pasquier is a professor of human resource management and labour relations, HEC Montréal. This column was originally published at The Conversation Canada: theconversation.com/ca.