Cisco, PwC and Wyndham show how companies actually retrain staff to work with AI

While part of the market debates how many jobs AI will take in the next few years, some large companies are betting on a different scenario: instead of cutting teams, they retrain them quickly for new processes. At the Semafor World Economy forum, executives from Cisco, PwC and Automation Anywhere explained how they are trying to build AI into their work without wrecking their workforce foundation.

Who is already training people

Corporate AI training still involves more talk than systematic practice. In the US, support measures are only taking shape: in February 2026, a bipartisan AI Workforce Training Act bill was introduced, with tax breaks for companies that train employees in prompt engineering, data literacy, machine learning and AI ethics. But while regulators argue over rules, the real burden has fallen on business, and a Gallup survey named manager support as the main factor in successful AI adoption.

PwC's chief AI officer Dan Priest offered one telling example. The company helped Wyndham deploy an agent-based system for handling customer requests, which cut call length by at least 30%. The freed-up time did not go toward layoffs: managers were able to retrain employees for tasks where human involvement matters more, such as higher-quality conversations with guests, oversight of the agents themselves, and handling non-standard situations.

"The goal was not to replace these people,"

Priest said, describing Wyndham's approach to automation.

PwC used similar logic in its work with Lucid Motors, where AI improved financial forecasting tools. The point was not to make fewer people necessary, but to move them to more valuable skills. Automation Anywhere head Mihir Shukla backed the same idea: effective adoption starts not with handing out trendy AI tools, but with rebuilding the workflows themselves and the human's role within them.

How the training works

Cisco's approach is tougher and more formal. According to Liz Centoni, the company's chief customer experience officer, a basic understanding of AI is now mandatory for everyone. She says 98% of Cisco employees already use AI tools daily, and the training is built as a hands-on program with an internal level system reminiscent of karate belts. By the company's design, this format should make skills measurable and eliminate training that exists only to satisfy reporting.

• Basic AI literacy is mandatory across the company
• Courses differ by role and depth rather than following one template for everyone
• Employees learn on real scenarios and workflows involving agents
• Once routine work is automated, people move to higher-value tasks
• Progress is tracked through clear levels and completed modules

There is a separate emphasis on the fact that different employees take to training differently. PwC noticed that younger staff respond better to short video instructions tied to specific tasks, while older, more experienced employees respond to in-person discussions about which non-technical skills become more important when AI takes over some functions. A one-size-fits-all program tends to hurt rather than help here, because the motivations and fears of the groups differ sharply.

Shukla adds another principle: people should be trained not alongside an abstract AI course but inside real work. His logic is that an employee should push a model to the point where it starts making mistakes, and at that moment understand what their own unique value is. This approach ties upskilling not to a formal checkbox in an LMS, but to career growth, accountability and a new role for the human inside an automated team.

Why junior hires matter

Amid claims that agent systems make junior positions unnecessary, Priest takes the opposite stance. In his view, companies still need to keep hiring entry-level staff, even if some starter tasks get automated. He describes the future structure not as a classic pyramid but as an hourglass: many entry-level employees at the bottom, less middle management in the center, and experts at the top who set the framework and control quality.

The logic is pragmatic. Juniors are easier to train for new processes, they are cheaper for the company, and they are often more willing to adopt AI. Sharp headcount cuts for quick efficiency, by contrast, can lead to lost institutional memory, lost oversight and expensive rehiring a year or two later. Besides, responsibility for an agent's mistakes still rests not with the agent but with the person who deployed and supervises it, especially in regulated industries, where the cost of failure is higher than usual.

What this means

The clearest corporate AI strategies right now look less like "we'll replace people with bots" and more like "we'll remove the routine and quickly retrain the team for new work." For the market this is an important signal: the winners will not be the companies that talk loudest about automation, but those that can turn it into skill growth rather than just a pretext for layoffs.