Follow ZDNET: Add us as a preferred source on Google.

ZDNET’s key takeaways

• Employees need structured training to prepare for AI.
• Companies should prioritize upskilling over layoffs. 
• Entry-level employees are still crucial to organizational health.

At Semafor World Economy this week, I watched moderator after moderator ask CEOs whether they think AI will eliminate jobs. While most CEOs responded with the augmentation argument — that AI is helping workers do more, not replacing them — new jobs, especially at the entry level, are dwindling. At the same time, AI could perform competitively on many work tasks by 2029. 

Beyond regulation, upskilling workers in AI is a key solution for mitigating these economic impacts. But where has upskilling been primarily a talking point, and where is it actually happening, meaningfully? 

Also: This AI expert says the job apocalypse isn’t coming, even if you’re a coder – here’s why  

I spoke with several leaders at Semafor’s summit about what they’re doing to retrain their employees for a very different near future, and what they find works best. 

Where upskilling is working 

Without initiatives from their companies, workers are left to navigate the task of upskilling on their own. On the policy side, the AI Workforce Training Act, a bipartisan bill introduced in February, proposes tax credits for companies that upskill their employees in prompt engineering, data literacy, machine learning, AI ethics, and other relevant topics. The Trump administration’s latest AI regulatory framework also calls for AI training and apprenticeships.

But until policy solidifies, companies themselves have the most potential for impact. A Gallup poll published Monday found that manager support is the primary driver of successful AI adoption. 

Also: Half of all US employees use AI at work now – and waste almost 8 hours a week doing it

Dan Priest, chief AI officer at PwC, develops AI strategies with various clients and said he’s seen a range of approaches to upskilling across the companies he works with, from formal to informal. Regardless of approach, he thinks effective leadership includes upskilling by default — it’s just good strategy. 

For one example, PwC helped hotel company Wyndham create an agentic system to handle customer requests, reducing call times by at least 30%. Employees learned how to oversee the agents themselves. That left managers much more time to train those employees in new skill areas, such as how to offer guests a better, more engaged experience. 

“The objective wasn’t to replace those people,” Priest emphasized, noting that Wyndham’s reinvestment in those workers was the key to the program’s success. 

Also: Turn AI chaos into a career opportunity by preparing for these 4 scenarios 

He oversaw a similar process for improving financial forecasting tools at Lucid Motors. In Priest’s telling, the employees involved weren’t utilized less or subject to layoffs because of the help agents provided; the initiative directly led to new skill development. 

Unlike for some of Priest’s clients, Cisco mandates AI upskilling, said EVP and chief customer experience officer Liz Centoni. 

“It’s a requirement for everyone across the organization because the thought is that there needs to be at least a core understanding of AI,” she said, noting that 98% of Cisco’s employees use AI tools every day. The company’s learning team has always maintained a program for both clients and its own employees, even prior to AI; that technical investment is core to how the company operates. For AI, that looks like hands-on training, complete with a karate-style system for assessing completion: employees can earn blue, white, and green belts for modules. 

Centoni noted that, because the jobs her teams perform will change as AI advances, Cisco’s approach to upskilling includes rethinking the work itself. 

“More than just bolting AI onto existing workflows, how does the work itself need to change?” she said. 

Also: AI is more likely to transform your job than replace it, Indeed finds 

Mihir Shukla, CEO of Automation Anywhere, agreed that successful upskilling needs a strategy for “how work should get done.” That means “delivering autonomous IT, autonomous supply chain, autonomous finance, autonomous claims, autonomous healthcare, etc., at a higher level than just sprinkling AI into the workforce or throwing tools at them,” he said. Successful approaches let employees learn by using actual agents in real workflows. 

“We recently challenged our engineering team to create software that was fully autonomous end-to-end, with no human coding or any employee who touched it,” Shukla explained. 

Adapt training for experience 

Internally, Priest has also seen success upskilling employees at PwC — but noted that avoiding a one-size-fits-all approach makes the difference. 

“Different generations in the workforce are going to respond differently to AI,” he said. “We made these short video explainers for doing some specific task with AI, and the younger hires really respond well to that; they like that format. But that doesn’t work for partners. For them, we’re bringing them into a room and discussing how to develop their non-technical skills in new ways based on what AI is handling now.” 

While Priest said most younger hires have demonstrated “exuberance” around learning to use AI, leaders need to be sensitive to how that openness varies based on time spent in the workforce. 

Also: Nervous about the job market? 5 ways to stand out in the age of AI 

“You’ve got people who’ve worked for 20 years to become specialists in their fields, and you’re telling them they have to change everything they’ve been doing now. That’s extremely hard,” he noted. “Leadership has to clarify what’s changing, but they also need to say what’s staying the same.” 

Priest advises companies to “pick their spot” when upskilling: focusing on the most urgent group of employees to implement AI first, based on their objectives. He isn’t concerned that this could leave certain types of work or employees behind. 

“I haven’t seen a category of job that’s being ignored,” he said of upskilling efforts, noting that if it’s happening right now, it’s only a temporary circumstance. 

Centoni said Cisco follows a similarly personalized approach. 

“By job category, there are courses on AI available depending on the role that you’re doing,” she explained. “The use cases are going to be different, so what level of depth do you go into?” 

Prioritize talent

Similarly to other arguments for AI’s positive potential for work, Centoni said that Cisco’s AI upskilling efforts so far have surfaced talent from employees with years of institutional knowledge that was otherwise buried in rote tasks. 

“Problems you cannot prompt your way through, that knowledge is just sitting in an engineer’s head,” she said. “When we used AI to automate all of the things that could be automated, that elevates these folks.” 

Seeing what AI has made room for in her employees, she’s now rethinking what to look for in future hires, and is focused on creating an environment for talent to flourish even more. Centoni doesn’t have an answer yet, but knows it will change the talent the company seeks and how they evaluate their success once onboarded. 

Also: Is AI coming for your job? Here’s one labor indicator that could soothe your fears

While Centoni was not prepared to say whether upskilling is mitigating layoffs at Cisco, her heightened interest in the kinds of people she still needs to hire was clear. Shukla takes a similar approach. 

“Every employee needs to operate in a place where you push a model to where it fails, and then you understand where your unique value lies,” he said. “For us, AI upskilling is directly tied to career growth: whether you stay at our company for decades or a few years, we feel a responsibility to cultivate a mindset around leveraging AI and mastering how to use it.” 

Priest shared that emphasis on talent, which is why he doesn’t think rapid layoffs are the answer to optimizing with AI. Companies getting results, he said, are turning to upskilling first. 

“It all centers around talent,” he said. “There’s a reason there’s a talent war in Silicon Valley right now.” Though similarly reticent to talk about layoffs, Shukla agreed that slapdash implementation sacrifices institutional knowledge and often means companies need to rebuild jobs at a higher cost later. 

It’s a valuable insight given how quickly some companies have cut staff in the name of AI. Anthropic CEO Dario Amodei recently upped the ante on his job elimination predictions. Just this week, Snap laid off 1,000 people, partially crediting the tech; over the last few months, major tech companies like Meta, Oracle, Block, and others have let more go. While it’s not clear from the outside whether these decisions reflect AI capabilities or simply correct bloat, the rhetoric could influence other, less AI-savvy sectors to follow suit. 

Keeping humans on staff is also crucial for navigating liability, Priest noted. “We’re a regulated business,” he said. “Something goes wrong — you don’t tell your agent, ‘Hey, you’re in trouble, you did a bad job,’ right? It’s the person who built that agent that’s accountable.” That system doesn’t function if companies cut huge amounts of staff and outsource tons of work to AI. 

Entry-level hires still matter 

Several Semafor summit attendees told me, with some hesitation, that agentic tools were making their junior hires appear obsolete. Contrary to a steady stream of research confirming entry-level jobs are falling, Priest said companies still need to hire for those positions, despite what they may think AI can automate. 

Also: The great AI skills disconnect – and how to fix it 

“Up until now, the structure of work has been based on a pyramid, right?” he said, referencing a traditional structure of many workers and fewer leaders at the top. “I see the ideal structure moving more toward an hourglass: you’ve got lots of entry-level hires on the bottom who you can and should invest in,” and fewer middle managers. Compared to highly motivated, upwardly mobile, and more affordable junior hires, he said, middle management can be harder to reskill. 

But ensuring workers don’t simply offload tons of work to AI is also central to investing in those younger hires, he clarified. 

“With senior employees, there’s maybe not enough adoption, and then with junior hires, there’s likely too much,” he said.