“It would not surprise me, if across all industries—whether it be front office, back office, legal, marketing, parts of finance—that you see 30% to 40% productivity improvements,” said Schulman of generative A.I. during a virtual conversation, hosted by Fortune, with a group of CEOs discussing emerging tech trends.

This spring, Schulman was one of the many CEOs to discuss A.I. during his latest earnings call, telling investors that the technology would enable PayPal to “meaningfully lower our costs for years to come.” A recent CEO survey conducted by Fortune and Deloitte showed that 79% of CEOs expect generative A.I. investments will increase efficiencies, with three out of four polled also saying A.I. investments can automate manual operations.  

One arena of productivity improvement that PayPal is currently experimenting with is the use of A.I. in code development. PayPal rolled out tool sets including Google’s Codey, AWS’s CodeWhisperer, and Microsoft’s Codist to over 50% of PayPal’s developers and is already seeing at least a 30% improvement in productivity.

“The speed and the efficiency in which these models can assist programmers and others is unlike anything I have ever seen,” said Schulman.

Naturally, these trends raise questions about the impact on jobs. Schulman advises that companies need to start thinking about those ramifications now. “I think we need to be very, very responsible in thinking about hiring plans, and not hiring lots and lots of people only for them to be laid off later,” he said. “You need to really think about where A.I. is going to impact your functional areas and in what way.”

The adoption of A.I. and other technologies like cloud computing across the private and public sectors has consistently been held back by concerns about security and risk, as well as the impact on the workforce. Companies must be measured in the right level of risk they are willing to accept, as well as ethical and quality-control concerns. 

“There’s no doubt that the pace of adoption is rapidly increasing,” said Jason Girzadas, CEO of Deloitte U.S.

But as it pertains to jobs, he said, the main question asked today is what mix of jobs will be augmented by A.I., versus those that are either created or eliminated. “That dynamic is being widely debated,” said Girzadas.

Mark Newman, president and CEO at Chemours, said the chemicals company’s development of new molecules is the area of the business that can benefit most from generative A.I. Newman said Chemours is still in the “early innings” of using generative A.I. to develop molecules that the company can process to meet demand, adhere to regulatory requirements, and are safe to produce.

To be more thoughtful about the opportunities that emerging tech like A.I. can create, Chemours recently hired a new chief enterprise transformation officer. 

“How do we take all the digital tools, including A.I., and drive massive productivity improvement in manufacturing, in our R&D center, and in our functions that support all that activity?” Newman said rhetorically. 

Educational tech company Skillsoft believes that every knowledge worker who isn’t already using A.I. today will utilize the technology in some form in the future. Skillsoft itself thinks of A.I. through three lenses: changing what it teaches, how it teaches, and how Skillsoft itself operates. 

“We are seeing that almost every customer conversation now starts with the need to prepare the workforce with this technology,” Skillsoft CEO Jeff Tarr said.

Debanjan Saha, CEO of enterprise A.I. platform DataRobot, outlined one use case they’ve built for A.I. that involves sorting through vendor proposals. What DataRobot has done is create a knowledge base of all past proposals that the company has answered, and so when a new proposal comes in, the predictive model is able to score each of the answers they are looking for. This process of sorting through RFPs at a quick and more efficient pace would be especially helpful for smaller and midsize companies that have lean legal teams.

But Brad Jackson, CEO of consulting firm Slalom, said his conversations with top executives show there is fear of early adoption of new A.I. technologies. No one wants to be first and risk a wave of negative news headlines if things go wrong. Jackson believes these fears might lead top executives to focus more on the efficiency benefits of A.I., while not yet fully exploring new value creation possibilities. 

Schulman answered this concern by saying A.I. needs to be rigorously tested, and applied in a controlled and responsible way. But ultimately, there is always some risk. 

“I would not have the fear of being first,” said Schulman. “Because you are already not first—I promise you that.”