The first years of my career in media were spent making photocopies, collecting faxes and filing papers. I also opened reader mail, took notes on meetings and ran errands for my bosses -- picking up everything from skim lattes to control-top Donna Karan tights.

I’ve been thinking back to those dues-paying days recently due to an onslaught of depressing research about how generative AI could hurt the career prospects of Generation Z, or those born 1997 to 2012. If it transforms or even takes over typical entry-level tasks in professional occupations -- building spreadsheets, writing first drafts or creating presentations -- what will happen to jobs for younger workers?

At first blush, the outlook is bleak. Long before generative AI was a buzzword, entry-level jobs were becoming tougher to land. The Great Recession caused a spike in the unemployment and underemployment rate for recent graduates. Younger workers have smaller networks, which makes job-searching tougher.

And many employers have simply decided they don’t want to bother with brand-new workers. Employers don’t want to invest time and money in training. That’s short-sighted, but it’s a trend entry-level workers have been fighting for decades.

Some say that AI will compound these problems. Not only will there be fewer professional jobs to go around, but as more of those tasks are automated, young workers will miss out on learning opportunities and fail to build the skills they need to advance.

It’s a fair concern, but it’s probably too gloomy.

“I don’t buy the doomsday scenarios at all,” says Amit Joshi, a professor who studies artificial intelligence at IMD business school in Lausanne, Switzerland. He says it’s more likely that AI will raise the benchmark for what an entry-level worker can achieve. Instead of taking a week to analyze and summarize 50 academic papers, a research assistant using ChatGPT might only need a day to synthesize the core points and check for errors.

It’s not only that a PowerPoint that used to take four hours to build might now be completed in an hour. It’s that the young workers who struggled with basic writing and coding will now have a tool that helps them get to a higher minimum standard.

If entry-level workers get that much faster and better, won’t companies be able to get by with fewer of them? Maybe. Those are among the reasons that studies from Pew and Brookings flag entry-level workers as particularly at risk of losing employment to AI.

But eventually, new jobs -- indeed, new industries -- will be invented.

After all, when photocopying and faxing disappeared as entry-level tasks, entry-level jobs didn’t disappear. Arguably, they got more interesting. If I were a new grad in journalism today, I would probably be a social media manager -- a job that didn’t exist when I graduated. And I might be better off: Writing social copy and making TikTok videos is undoubtedly more challenging and rewarding than fixing paper jams.

That’s not to say that there’s no reason to be concerned. Manufacturing provides an obvious example of an industry where automation destroyed good blue-collar jobs. Some experts predict that generative AI could have a similar effect on computer-based jobs for workers without college degrees, like customer service.

Among the college-educated workforce, the advent of generative AI is less likely to erase jobs and more likely to prompt employees to learn new skills. For example, if a bot can churn out fluid-sounding text in minutes or even seconds, perhaps the human’s job shifts from writing to editing -- checking the facts to make sure they’re correct, finding ways to improve the logic and persuasiveness of the piece, injecting verve and style. All of this requires critical thinking.

Generative AI still needs human oversight -- and maybe always will, despite its ability to rapidly learn new skills. As Walmart executives put it in a LinkedIn post recently when the company announced the rollout of a generative AI app for its office employees, “GenAI can help us work faster and more efficiently, but it also has limitations: It lacks judgment, has a limited understanding of context and is only as good as the data it’s trained on.”

For an example, consider my own early experiment testing ChatGPT.

To see whether it could help me chase down a statistic for a column on child care, I asked it how many remote workers had kids at home without day care. It responded with a statistic and cited FlexJobs; when I asked for the URL, it provided one. When that produced only an error page, I contacted FlexJobs, and they said they could not remember ever conducting such a survey. It seems ChatGPT just fabricated it. A hallucinating assistant is far worse than a slow one.

The college graduates of the future will still be able to find jobs, even if the hunt is grueling -- and even if we can’t predict exactly what the work will look like. I just hope it’s more interesting than buying someone else’s pantyhose.

Sarah Green Carmichael

Sarah Green Carmichael is a Bloomberg Opinion editor. -- Ed.

(Tribune Content Agency)