Pew Research Center recently published interesting results from a large survey about AI usage in the workplace. In late 2024, they surveyed over 5,000 adults employed in the U.S., with the sample weighted to be representative of the overall U.S. adult population. Here’s what the findings mean for designers working on AI features.

Most US Workers Are Not Using AI

In 2023, after ChatGPT became publicly available, we were flooded with predictions of how AI would revolutionize workplaces in a few short years. Two years later, that future has yet to materialize.

In Pew’s survey, 81% of the workers sampled were considered non-AI users. These respondents said little or none of their work is done with AI. 17% of workers sampled said they hadn’t even heard of AI being used in their workplace.

It’s unclear which AI use cases respondents were asked about in the survey, but it included questions about AI chatbots, such as ChatGPT, Gemini, or Copilot. 55% of workers say they rarely or never use AI chatbots at work, and 29% of workers haven’t heard of these products! 

You might wonder whether workplaces prevent workers from using AI chatbots, but this doesn’t seem true. Only 9% of respondents who don’t use AI chatbots say their employer does not allow them.

Workers’ low adoption rate of AI is unsurprising since the survey included industries where AI is less prevalent (such as the services sector or construction). Over a third of users (36%) who don’t use AI chatbots said there wasn’t a use case for them in their work. 

Tech Workers (Like Us) Are More Likely to Use AI

Pew’s research found that workers adopting AI are more likely to work in jobs involving data processing (63%), information technology (12%), as well as industries like finance, real estate, and insurance. These professions are more likely to introduce AI in the workplace, provide training, or have work duties that would benefit from AI. Information technology workers were more likely to report that their employer actively encouraged using AI (36% of respondents).

AI adopters at work are also more likely to live in urban areas and have a bachelor’s degree (51% versus 39%). Where someone lives and whether they have a bachelor’s degree are naturally connected to their work type.

You Probably Know More About AI Than Your Users

If you’re reading this article, you are likely using AI more than the average American worker.

As designers, we may assume that our users have similar skills and familiarity with AI as we do — when they don’t. This is the false-consensus effect in action — the human tendency to assume others will think and behave like we do. 

This phenomenon has been responsible for many design failures. If you’re a highly tech-savvy person who works and socializes with similar people, you may be in a techy “bubble.” It may feel like everyone you know is using genAI frequently, so you subconsciously assume your users do too. However, depending on your audience, Pew’s findings suggest that extrapolation may be wrong.

Younger Workers Use AI More

AI adopters at work are more likely to be younger than non-AI adopters (73% of AI adopters were under 50 versus 65% of non-adopters). Workers aged 18–29 were the most likely to report using AI chatbots at work out of any other age group. 

Additionally, younger AI adopters were more likely to use AI in a broader selection of work tasks than older AI adopters — and in more advanced ways, like ideation or computer programming. 

When analyzing the difference between the 18–49 age group and workers above 50, a higher proportion of younger workers reported using chatbots in almost all the tasks they were asked about than the older age group. These differences were statistically significant, except for information-seeking tasks (like researching a topic).

These findings are not surprising and follow a historic trend of young early adopters of new technologies. We’ve seen faster adoption of the internet by Millennials (“digital natives”) and faster adoption of social media by GenZ (“social media natives”). We’ll likely see even higher adoption for those growing up with AI (for posterity, we’ll call them “AI natives”).

Younger people tend to adapt to new technology more quickly. While the young workers in this survey didn’t grow up with AI, they may find it easier to pick up and adapt to this new interaction dynamic. Using genAI (especially in a broad, chat-based context) requires users to shift mental models about how they use digital technology.

Most AI Usage Is Not Sophisticated

The survey also asked how workers are using AI chatbots. The most-reported use case was information seeking (57%). This is consistent with our longitudinal research on using AI chatbots, where we found that information-seeking represented 75% of all recorded conversations.

Other common tasks among workers using AI chatbots included editing (52%) and drafting content (47%). Fewer workers said they used AI for more complex tasks, such as: 

• Ideation (35%)
• Analysis or computer coding support (27%)
• Creating and editing imagery or videos (21%)

Perception of Usefulness Increases as Usage Increases

Respondents were asked about how often they used AI chatbots at work. Not surprisingly, workers using AI chatbots frequently (at least a few times a month) rated their helpfulness higher than those who use AI chatbots infrequently (54% versus 25%, respectively). 

AI chatbots are not always forthcoming about their capabilities. In our early qualitative research of how users engage with AI chatbots, many of our users (particularly people unfamiliar with AI chatbots) were unsure what the chatbot could do. We observed many users utilizing Can You prompts to learn what a chatbot could help them with. It can take a while before users wrap their heads around how genAI works, where it can help in their day-to-day work, and how to get results through successful prompt engineering. 

Two years later, our studies with consumers show that people are still struggling to adapt their mental models to this new interaction paradigm, even those who use genAI fairly often. This is why narrowly-scoped AI features that serve a single purpose (like providing product recommendations or making a playlist) tend to be easier for new users to understand than open-ended, broad-scope AI chats.

As users become more familiar with and gain more practice, they become more competent at simple tasks (like information seeking) and begin to explore using AI for more complex tasks. As competence grows in using a tool, so does the realm of potential use cases and the perceived utility of a tool.

We’re still in the early stages of user competence with genAI. Researchers asked whether respondents believed chatbots helped them speed up their work and whether it improved their work quality. 40% of AI chatbot adopters believe these tools help speed up their work, but a smaller proportion of the same group (29%) believe they help improve the quality of their work. We predict that more workers will benefit from improved work outputs as competence grows.

Takeaways for Design Teams

Generative AI tools are being used for work, but not as pervasively as optimists predicted in 2023. New LLM models continue to surpass each other in benchmark testing, but that has yet to translate into meaningful impact on a large scale. 

Don’t expect people using AI features to be very familiar with them. Most people don’t use them regularly or at all. If they use them, they’re likely using them in very simple ways, such as to look up information. Remember: You are not the user. You not only have knowledge about your product that your users may not have, but you’re also probably more familiar with AI than they are. On top of that, you likely work at a company that supports or encourages using AI.

Younger workers are using AI more than everyone else. If you’re designing an AI feature (whether it’s a product for the general population or professionals), recruit across all age categories. Favoring only younger recruits (who tend to represent most people in internet-based recruitment panels and platforms) can skew your research findings. Testing with older users and users who might be employed in different industries will help you learn how to build a product that will be more intuitive and usable by people with less AI familiarity and experience.

People’s uses of AI are still pretty limited. As designers, we can’t force users to use AI, but we can make its benefits and use cases more transparent. Smart AI onboarding, relevant prompt controls, and progressively surfacing new use cases in AI features and products will be essential to support AI-skill growth and adoption.

References

Lin, L. and Parker, K. (2025) U.S. workers are more worried than hopeful about future AI use in the Workplace, Pew Research Center. Available at: https://www.pewresearch.org/social-trends/2025/02/25/u-s-workers-are-more-worried-than-hopeful-about-future-ai-use-in-the-workplace/ (Accessed: 27 February 2025).