For many Americans, AI is rapidly changing the way we work.

A growing number of workers now use AI at their jobs with some frequency. According to a recent Gallup poll, 40% of U.S. workers say that they use AI at work at least a few times a year, and 19% of workers use it several times a week. Both statistics have nearly doubled since last year, from 21% and 11%, respectively.

At the same time, over half of American workers are worried about AI’s impact on the workforce, according to a Pew Research Center survey. Their fears have merit: a World Economic Forum report published in January found that 48% of U.S. employers plan to reduce their workforce due to AI.

Naturally, the rapid growth of AI in the workplace has raised plenty of questions. How will AI reshape our jobs? What new skills will we need to develop? Which industries will be impacted the most by AI?

These questions don’t have easy answers, says Ethan Mollick, an associate professor at Wharton and author of “Co-Intelligence: Living and Working with AI.”

Mollick, who is also the co-director of Wharton’s Generative AI Labs, is well aware of concerns about AI replacing human jobs.

“The idea that you could just sub in AI for people seems naive to me,” he says. Still, as AI keeps improving, “there may be effects” for workers, he says.

Here’s what Mollick has to say about AI and the future of work.

CNBC Make It: There’s a lot of concern about AI replacing human jobs, including some big predictions from leaders like Bill Gates. What’s your take on that?

AI agents are not there yet. Right now, AI is good at some stuff, bad at some stuff, but it doesn’t substitute well for human jobs, overall.

It does some things quite well, but the goal of the labs is [to create] fully autonomous agents and machines smarter than human in the next 3 years. Do we know they can achieve it? We don’t, but that is their bet. That’s what they’re aiming for. They are expecting and aiming for mass unemployment. That is what they keep telling us to prepare for.

As for believing them or not, we just don’t know, right? You have to take it as at least a possibility, but we’re not there yet, either. A lot of it is also the choice of organizational leaders who get to decide how these systems are actually used, and organizational change is slower than all the labs and tech people think.

A lot of the time, technology creates new jobs. That’s possible, too. We just don’t know the answer.

As AI usage becomes more prevalent, what skills will we need to develop in the workforce?

If you asked about AI skills a year ago, I would have said prompting skills. That doesn’t matter as much anymore. We’ve been doing a lot of research, and it turns out that the prompts just don’t matter the way they used to.

So, you know, what does that leave us with? Well, judgment, taste, deep experience and knowledge. But you have to build those in some ways despite AI, rather than with their help.

Having curiosity and agency also helps, but these are not really skills. I don’t think using AI is going to be the hard thing for most people.

What is the “hard thing,” then?

I think it’s developing enough expertise to be able to oversee these systems.

Expertise is gained by apprenticeship, which means doing some AI-level work [tasks that current AI models can do easily] over and over again, so you learn how to do something right. Why would anyone ever do that again? And that becomes a real challenge. We have to figure out how to solve that with a mix of education and training.

How do you think AI will affect the entry-level job market?

I think people are jumping to the conclusion that [AI is] why we’re seeing youth unemployment. I don’t think that’s the issue yet, but I think that’s a huge concern.

Companies are going to have to view entry level jobs in some ways, not just as getting work done, but as a chance to get people who will become senior employees, and train them up to be that way, which is very different than how they viewed the work before.

Are your students concerned about AI’s impact on jobs?

I think everybody’s worrying about it, right? Consulting and banking, analyst roles and marketing roles — those are all jobs touched by AI. The more educated you are, the more highly paid you are, the more your job overlaps with AI.

So I think everyone’s very concerned and I don’t have easy answers for them. The advice I tend to give people is to pick jobs that have as many ‘bundled’ tasks as possible.

Think about doctors. You have a job where someone’s supposed to be good at empathy and [surgical] hand skills and diagnosis and be able to run an office and keep up with the latest side of research. If AI helps you with some of those things, that’s not a disaster.

If AI can do one or two of those things better than you, that doesn’t destroy your job, it changes what you do, and hopefully it lets you focus on the things you like best.

So bundled jobs are more likely to be flexible than single thread jobs.

How might AI adoption play out in the workplace?

For me, the issue is that these tools are not really built as productivity tools. They’re built as chatbots, so they work really well at the individual level, but that doesn’t translate into something that can be stamped out across the entire team very easily.

People are still figuring out how to operate with these things as teams. Do you bring it into every meeting and ask the AI questions in the middle of each meeting? Does everybody have their own AI campaign they’re talking to?

The piece I keep making a big deal about is that it is unfair to ask employees to figure it out. I’m seeing leadership and organizations say it’s urgent to use AI, people will be fired without it, and then they have no articulation about what the future looks like.

I want to hammer that point home, which is, without articulating a vision, where do we go? And that’s the missing piece. It’s not just up to everybody to figure it out.

Instructors and college professors need to take an active role in shaping how AI is used. Leaders of organizations need to take an active role in shaping how AI is used. It can’t just be, ‘everyone figure it out and magic will happen.’

This interview has been condensed and edited for clarity.

Want to stand out, grow your network, and get more job opportunities? Sign up for Smarter by CNBC Make It’s new online course, How to Build a Standout Personal Brand: Online, In Person, and At Work. Learn from three expert instructors how to showcase your skills, build a stellar reputation, and create a digital presence that AI can’t replicate. Sign up today with coupon code EARLYBIRD for an introductory discount of 30% off the regular course price of $67 (plus tax). Offer valid July 22, 2025, through September 2, 2025.