Ethan Mollick is considered one of the top experts on the emerging world of generative AI, jobs and the economy, but you might not think so if you take his word for it.
“Leading is not that far a lead as an expert,” he told CNBC’s Sharon Epperson at the CNBC Workforce Executive Council Summit in New York City on Tuesday.
The Wharton School of the University of Pennsylvania professor, who says he has counseled everyone from Federal Reserve Chair Jerome Powell to Jimmy Fallon on the new world of gen AI, shared a blunt message with human resources officers in attendance at the CNBC event. “I can tell you, no one knows anything,” Mollick said.
That includes, he says, the top AI research labs when it comes to the job market and use cases. “They don’t know what it’s useful for. They tell me they use my Twitter [X] feed to figure out use cases,” he said.
Underlying his words was a simple point: No firm can hire a worker today who has five years of experience using gen AI. “They don’t exist,” Mollick said.
To be sure, there is some evidence emerging of workplace productivity gains from gen AI, and Mollick and chief human resources officers at the CNBC WEC Summit shared some evidence from their research and real-world experience with workers, at companies from Walmart to Verizon, and JPMorgan. But there was general agreement that there remain more questions than answers today for corporate leaders when it comes to AI and the workplace.
“We frankly don’t know what the future looks like,” said Claire MacIntyre, Sam’s Club senior vice president and chief people officer, in a separate Summit session with CNBC’s Morgan Brennan. “This is the worst version of the technology we will ever use,” she said.
Need to shift away from rewards based on having all the answers
Much of the progress within AI operates within a realm that tech experts describe as a “black box,” and experts at the CNBC event said there is a comparable gap that exists today in our understanding of AI’s impact on the economy that spans from early education through professional careers.
MacIntyre said modern career culture is predicated on being rewarded for “having answers” and that is a process that began in the education system. But that is shifting for leadership and workers. Leadership, in particular, she says, “is no longer about having answers. It is actually now about asking brilliant questions, editing information and making decisions at the speed of TikTok,” she added.
Christina Schelling, Verizon chief talent officer, who spoke on the same panel with the Sam’s Club executive, agreed. For decades, she said, “We were rewarded for perfection and being an overachieving perfectionist in the workforce.”
But with AI, Schelling says, “the outcome is rarely perfect or the one you need exactly to move forward. It’s ok to be ok with failing or being incorrect,” she said. Now, how quickly you can rebound and continue to test and try new things is as likely to be the successful model as the way we have been rewarded since kindergarten, she said, even though it runs counter to it.
“What we are trying to focus on is less learning as action, but more as a mindset,” MacIntyre said. “Be curious and be able to unlearn, and be feedback-literate.” All of this, she says, is critical to how culture needs to evolve.
For employers, that makes hiring a more difficult equation, according to Kiersten Barnet, executive director of the New York Jobs CEO Council, which was started by JPMorgan Chase CEO Jamie Dimon and other CEOs of the largest employers in the city. “Everyone knows we will need something a little different from before, but we don’t know what that will look like in five, ten years,” she told CNBC’s Brennan in a one-on-one interview at the Summit.
She also drew a direct line to education, a focus for her organization, which is working with colleges and high schools in New York City to prepare workers for jobs that will require AI skills to build solid career pathways and earning potential. “Think about traditional classrooms. They look the same as 100 years ago as far as the way we learn. Even if the content is different, you don’t learn critical thinking from a textbook,” she said.
She noted that the New York Jobs CEO Council is involved in an effort to make gen AI a requirement for students, and OpenAI is working on certifications which she thinks will be embedded quickly and adopted in coursework and ultimately lead more work roles to be thought about in terms of applied use of AI technology, but she added that it remains an “if.”
“We don’t have it right now and it is hard to assess everyone’s ability on the applied side of the technology,” she said.
What we do know about AI, workers and jobs
Barnet said there are some bets she is willing to make on what will work for workers in the future. For one, the ability to flexibly and continuously learn “is a skill in and of itself,” she said.
Softer skills are more important than ever, she added, specifically because of “the uncertainty of the future” and knowing some skills AI can do for us.
Schelling stressed that it has long been known that empathy, curiosity, agility and decision-making skills are all important to success. but they are going to be more heavily weighted now and factored into a more complex job market in an AI world. It is already a data input in hiring and career advancement, but at the same it is also becoming “something largely unknown or new, so the gray takes on a bit more meaning,” she said.
Mollick says logically this makes sense, because current AI is much more like a human than a machine, so people who are good with people can use it to succeed.
He also pointed to evidence from a study he worked on with Boston Consulting Group that showed significant improvements in work productivity from the use of gen AI, as well as a study from Procter & Gamble that found employees performed as well as teams when assisted by AI.
“We know the impact is there,” Mollick said, but he stressed that when it comes to fears of job replacement, he sees it as a choice leadership will face, and execute badly. “I worry without imagination, organizations will think automation is the way to go,” Mollick said. And he said in the current environment, workers will be reluctant to embrace AI if they feel like the productivity gains won’t come back to them in the form of additional benefits.
Companies, including Sam’s Club and Verizon, are already seeing results today from early adoption. At the Walmart company, over 100,000 frontline workers have used gen AI over the past 18 months, including frontline managers using ChatGPT to help them run their businesses, as well as computer vision on autonomous scrubbers going around and doing inventory counts and other mundane tasks that associates can now skip.
At Verizon, there is also a focus on the frontline workers that directly interact with customers, but Schelling said the company has reached the stage of moving from pilots to “full enterprise transformation … an AI overlay to the company.”
One of the biggest projects at Verizon was using gen AI to scour all publicly available information on the company’s more than 100,000 employees to build a better AI system for match workers with potential career pathways. The company’s AI was able to clean up its data on roles and skills to identify career pathways in the abstract, but couldn’t match it to the actual workers without more complete information on their lives.
“We didn’t have enough data on employees,” Schelling explained. “We found they are more likely to update external than internal profiles. So we pulled every available public piece of information on employees with AI and fused it with internal employee profiles,” she added.
Employees were part of the process — though they had to opt-out rather than opt-in — and they were asked to change and modify information if inaccurate. Ultimately, Verizon went from less than 5% complete data sets to close to 100%, and it is working to the benefit of employees — nudging them with jobs that can fit based on their skills, as well as suggestions for training and certifications that help lay out a job they want “10 years into the future,” Schelling said.
While workers were hesitant at first about the fusion of the external and internal data, she says it is seen as a value-add, including less than a 1% attrition rate in the pilot group.
No. 1 piece of advice that costs $20 a month
Mollick had three structural pillars to suggest for organizations to move ahead in a constructive way: developing AI in leadership, creating an AI lab, and getting AI out to the crowd.
And it is all changing very quickly. “Almost everything we knew about training people doesn’t apply anymore. None of the prompting from four months ago works,” he said. “Prompt engineering doesn’t matter anymore. Saying the right words or being nice doesn’t matter, but giving it context we give to humans to make decisions does matter,” Mollick said. “You need to ‘crowd’ the best AI users and take ideas from the crowd and turn them into products that people use right away,” he added.
And there is only one way to start doing that, according to Mollick. “My No. 1 piece of advice is to pay $20 a month for [Anthropic’s] Claude or [OpenAI’s] GPT or [Google’s] Gemini and use it for everything you can use it for legally.”
Mollick says to use AI for 10 hours minimum a week. “It’s not that hard,” he said, and you will quickly learn what it is good at and what it isn’t good at. “You can’t push it down. You have to use it yourself as a leader. You can’t say you will set up time to do it,” he added.
As for all the vendors selling tools, he says most are just reselling GTP, Gemini or Claude and have no better access to AI than anyone else does. “I can’t tell you and no one can tell you unless your lab tries it,” he said.
“Let everyone ‘do’ and some people will be good off the bat and they become the lab and innovation,” Mollick said. “Waiting for answers or letting IT take it over are the biggest mistakes HR leaders can make,” he added. “As soon as you turn them onto tools, you can figure out use cases.”
To join the CNBC Workforce Executive Council, apply at cnbccouncils.com/wec.