3 Studies Give a Window Into How AI Is Reshaping the Workforce

It’s been over two years since ChatGPT was introduced to the world — and the workplace is still figuring out how to adapt to AI.

One large-scale field experiment and two surveys published or conducted last month give some clues into how AI is reshaping work.

While AI has shown signs of helping close experience gaps and improve collaboration, the data suggests some workers are skeptical and cautious — particularly when guidance on how to use AI is limited or unclear.

Here’s what researchers found.

AI can be a useful teammate

Workers using AI performed just as well as two-person human teams, and AI-augmented teams were significantly more likely to produce top-tier solutions, a working paper has suggested.

In a pre-registered, randomized controlled trial involving 776 professionals at Procter & Gamble, researchers from Harvard, Wharton, and the Digital Data Design Institute assigned participants to complete real-world product development tasks. They included generating packaging ideas and proposing retail strategies, either with or without the help of GPT-4 or GPT-4o.

Participants were randomly placed into one of four groups: working solo or in teams, and with or without AI access. The study measured solution quality, task completion time, and emotional response.

Individuals using AI performed as well as human teams working without AI, the study found, and AI-assisted teams performed best overall and were more likely to generate the top 10% of solutions than those working alone.

Ethan Mollick, a professor at UPenn’s Wharton School and one of the paper’s authors, said in a March blog post the findings suggest “AI sometimes functions more like a teammate than a tool” and that this should make companies think differently about the technology.

He said that businesses should rethink how they structure teams, train workers, and assign cross-functional tasks.

The study also found that when given access to AI, employees with less product development experience performed at levels comparable to veteran teams, suggesting the technology can help close functional knowledge gaps.

Working with AI improved the emotional experience of work, the study suggested. Participants using AI reported higher levels of excitement, energy, and enthusiasm and lower levels of anxiety and frustration compared to those working without it.

In some cases, individuals working with AI had more positive emotional responses than those working in human teams.

While the experiment involved just one day of work, the paper is still a working draft and hasn’t yet been peer-reviewed. The researchers acknowledge caveats: the study involved only small teams participating in just one day of tasks, and relied on GPT-4 models.

Most Gen Z workers say they lack AI guidance

Even as digital natives, most Gen Z workers are navigating AI at work without clear guidance, a March survey from Gallup and the Walton Family Foundation suggested.

More than half of Gen Z workers — 55% — said their employers have no formal policy on AI use. And among those whose workplaces do have rules, only about one in 10 describe those policies as “extremely clear.”

The Gallup web survey of 3,465 13-to 28-year-old Americans across all 50 US states said that Gen Z students and employees in environments with clear AI guidelines are more likely to use AI regularly than their peers.

At the same time, many younger workers are skeptical about the quality of AI-generated work. Nearly two-thirds of Gen Z respondents said they’d be more likely to trust human-produced work compared to work done or assisted by AI.

The findings paint a conflicted picture of how Gen Z as a whole perceives AI’s impact on their critical thinking versus their efficiency at work or school. Nearly half — 49% — of Gen Zers thought AI would harm their critical thinking ability, while just 22% said they thought it would help.

Yet many reported practical benefits, with 72% saying AI can help them find information more easily and 66% saying it can help them work faster.

Software engineers are split over AI

Vibe coding — using AI to create software using prompts — is shaking up software development. However, not all software engineers appear to be on board with the movement.

In a Wired survey of 730 software engineers published in March, three in four developers said they had tried AI tools like ChatGPT, and most of them said they used AI at least weekly.

But enthusiasm varied. Nearly 40% identified as “AI pessimists,” while just over a third said they were optimists. Mid-career coders were the most skeptical group, with almost half saying they were pessimistic about AI’s impact.

Meanwhile, early-career devs appeared to be more optimistic. Three-quarters of engineers who had been coding for less than a year said they were AI optimists. And nearly one in three veteran coders (with 20+ years’ experience) said they had already integrated AI into their workflows.

But not all adoption is out in the open. About 4% of full-time programmers said they used AI at work without telling their employer.

That unease might reflect broader job market worries. On Indeed, US software engineer job postings are down a third compared with five years ago.