Uppsala researcher says AI’s effect on work engagement depends on job design

Uppsala University researcher Andreas Bergqvist says AI can strengthen or weaken engagement at work depending on how it changes people’s jobs, in a university interview published by Phys.org on October 9. His team is developing a framework to help employers, unions and developers identify conditions that support meaningful, manageable work.

The interview describes research across IT, agriculture and rail transport. Bergqvist, a doctoral student in Uppsala’s Department of Information Technology, says introducing AI, automation or robotics does not automatically increase productivity or engagement. The effects depend on how technology interacts with people and the organization of their work.

How automation changes meaningful work

In the university interview, Bergqvist describes changes extending beyond individual tasks: technology can redistribute responsibilities, alter the skills a job requires and change how much influence employees have over their work. Removing simpler tasks can also increase the intensity of the work that remains.

That helps explain why the same technology can produce both engagement and frustration, he says. The relevant question is whether employees can still perform work they regard as meaningful, manageable and conducive to development, rather than simply whether a workplace has introduced a new system.

“Employees work with technology, not alongside it,” he says, describing systems that increasingly shape how work is organized, prioritized and carried out. This is his interpretation of the research across the three sectors, rather than a measured result for the economy as a whole.

His recommendations emphasize freeing time for skilled tasks, giving employees opportunities to use their abilities and preserving their influence over how work is performed. These are conditions the researchers associate with stronger engagement, not guarantees that any particular AI installation will improve a workplace.

What earlier OECD workplace surveys found

Independent historical evidence comes from the OECD’s Employment Outlook 2023. Its analysis draws on surveys conducted in early 2022 involving 5,334 workers and 2,053 firms in finance and manufacturing across Austria, Canada, France, Germany, Ireland, Britain and the United States.

The OECD reported that 63% of surveyed AI users said the technology improved their enjoyment of work. That figure records respondents’ perceptions. It does not establish an independently observed improvement in engagement or demonstrate that AI caused an increase in productivity.

Responses about autonomy also varied. Among AI users in finance, 58% said AI increased their control over the sequence of tasks, while 20% said it reduced that control. In manufacturing, the corresponding shares were 59% reporting increased control and 21% reporting reduced control.

The OECD also found that workers subject to algorithmic management were less positive about several aspects of job quality than other groups interacting with AI. Its analysis identified risks involving work intensity, autonomy and privacy, alongside the benefits workers reported.

Those findings provide context rather than a response to Bergqvist’s interview. They concern different sectors and an earlier period, and cannot be read as measurements of workplaces in October 2026. The OECD also said more research was needed to establish whether its findings extended to newer applications such as large language models.

A further limitation concerns whose experiences were captured: the OECD cautioned that results often describe workers who remained at firms adopting AI, potentially excluding people who had been displaced. Positive responses among those surveyed therefore do not account for every possible employment outcome.

What remains unknown about the Uppsala framework

Bergqvist says the framework under development is intended for researchers, employers, labor unions and systems developers. It aims to help them identify factors that could strengthen or weaken engagement in digital workplaces. The interview gives no completion date or results from testing a finished framework.

The interview also provides no full sample description, fieldwork dates, effect sizes or detailed methodology covering the three sectors. October 9 dates the published account, not the underlying research. Its claims remain Bergqvist’s attributed account of the work, without enough detail to quantify the effects or establish how widely they apply.