Good morning. Generative AI usage has continued to rise this year, even as sentiment within companies has cooled, a new study from the Wharton School of the University of Pennsylvania and research and analytics firm GBK Collective shows.

The comprehensive analysis of AI usage in the enterprise, now in its fourth year, included 1,416 U.S. enterprise leaders and entry- and mid-level employees.

“You could see a very marked AI backlash over the last months,” co-author Stefano Puntoni, Sebastian S. Kresge professor of marketing at the Wharton School and faculty co-director of Wharton Human-AI Research, tells the WSJ Leadership Institute. “But certainly, we see that generally people are positive,” he added.

Daily AI usage among respondents ranked manager or above rose 16 percentage points from last year to 62%. But that same group reported increasing caution and even skepticism about the technology.

That fear was particularly prevalent among some of the technology’s heaviest users. Eighty-two percent of respondents ranked vice president or higher say their focus has shifted to managing AI output. Almost half—48%—of the same tier worry about their own skills or thinking declining, compared with 32% of managers and directors.

The report reflects a growing maturity in expectations around AI, as organizations navigate concerns over costs and the increasingly autonomous agents, and as the definition of ROI begins to shift from time saved to the creation of broader value. 

With adoption largely settled, the harder task now is getting the whole organization moving in the same direction. The companies seeing more significant financial returns, the authors found, are those that have integrated AI most fully up and down the business.

A gap between the C-suite and cubicle

Setting an AI strategy at the top is only half the job. For it to stick, employees need clear answers to practical questions, including what rules govern its use, and who takes responsibility when it makes a mistake. Many companies have yet to close that distance between executive vision and everyday practice.

One of the report’s most consistent findings is how differently leaders and employees see AI’s progress.

Executives generally believe AI is transforming work faster than their employees do. On training, 78% of those who are managers and above say employees are adequately trained, but only 56% of the entry- and mid-level employees agree.

Co-author Jeremy Korst, a partner with GBK Collective and lecturer at the Wharton School, told the WSJLI that the gap matters because the middle of the organization is “where the work gets done.” Those mid-level managers and directors are now being asked to rethink workflows and function across organizational boundaries, usually without extra time or budget. 

More highlights from the report below:

ROI

Nearly all companies report some return on their AI investments, but far fewer report substantial ones: 80% of managers and above see positive returns, while 28% call them strongly positive. Part of the difference may come down to what companies measure. Hours saved is the most common yardstick, yet departments that judge AI by growth are far more likely to report strong returns.

“If you’re using it for growth, you have to start doing things that you were not doing,” said Puntoni. “And so it takes imagination, and I think that’s much, much harder.” 

His advice to executives: “Your job should not be to use this technology to automate what you’re doing today. It should be to figure out what it is that you could be automating that you’re not doing at all.”

Agents and responsibility

Companies are handing AI more autonomy. Two-thirds of departments now use AI agents in some form, and 17% let them act without a person approving each step. Returns on AI tend to be strongest in those departments using agents. 

“It seems to be going very fast,” said Puntoni about agent adoption. “Those organizations that are kind of stuck in the middle, that are adopting, but not quite, they seem to be struggling.”

But who answers when an agent goes wrong remains unsettled.

Companies doing it best

Departments with strong ROI were no more likely than others to reduce staff or freeze hiring. Instead, according to the report, they were more likely to have “changed their management structure (64% vs. 47%), significantly changed workflows (84% against 69%), and are more likely to monitor AI use (92% vs. 74%),” among other factors.

“This technology is not going to work by itself,” Puntoni said. “You have to figure out what to do with it.”

Tech-leader takeaway

Leadership matters as well. At organizations with a newly created chief AI officer, 58% of managers and above report strongly positive ROI, compared with 24% where the responsibilities were added to an executive’s duties and 8% where no such role exists.

Puntoni cautioned that the data can’t show whether the role drives results or whether companies that create the CAIO role were more AI-savvy to begin with. He suspects it runs both ways: Companies that did a good job of setting up the role probably already had the culture and capabilities to exploit the technology. “But having that role hopefully will help,” he said.

• Meta Platforms is looking to its just-released AI model, Muse Wonder, to power the subscriber versions of its Muse personal assistant, which has seen a decline of retention since its buzzy September debut, WSJ reports. Meta says Muse Wonder includes better memory and more expertise in various fields and performs better than OpenAI’s Astra and Anthropic’s Fable on several tests.

• Chip makers fell sharply in premarket trade after Chinese chip maker CXMT said it achieved a major breakthrough in memory technology that would allow it to produce more efficient chips, WSJ reports. Micron Technology and Sandisk dropped 4% and 5.2%, respectively, premarket, while Nvidia shares fell 1%. Tech struggled more broadly, with Nasdaq futures down 0.7%.

• Microsoft next year plans to deploy its latest AI chip, the Information reports. CEO Satya Nadella in a tweet said the chips, called Maia 3oo, deliver “40% more tokens per watt compared to leading GPU-based systems.”

• Hackers accessed cooling systems at hundreds of supermarkets in France on Sunday, heating refrigerators and freezers to 122F for around six hours, and leading businesses to throw out damaged food, Bloomberg reports. It wasn’t immediately clear who was behind the refrigeration cyberattacks.

The WSJ Tech Council brings together CIOs, CTOs and CISOs advancing innovation and shaping the future. Join this trusted community where tech executives connect with peers to explore emerging trends and gain the perspective they need to stay ahead of disruption.

Request Information

Follow Isabelle Bousquette on LinkedIn, Instagram, X, and