Startups in China can now develop digital avatars using just a few minutes of footage, with the results being nearly as good as a human streamer, Rodney Zemmel, senior partner at McKinsey & Company, explained during a lunchtime session on AI’s potential at the Fortune Global Forum in New York. (The session was sponsored by McKinsey).

“They didn’t outperform the best human live stream sellers, but they were outperforming the median human live stream sellers pretty quickly,” Zemmel noted. He pointed out that with the introduction of this AI, average wages of human streamers dropped by about 20%. “This is the first example you can see of robots outcompeting the humans in real time,” Zemmel said.

Some of China’s top livestreamers are now using AI to stream around the clock, using a digital avatar to keep selling products while the human behind it goes to bed.

Fellow panelist Barbara Humpton, chief executive officer of Siemens USA, said that AI can potentially help alleviate a tight labor market.

“We are in a labor constrained environment. Our workforces all around the world are shrinking,” she said. “The question is: Who’s going to do all the things that need to be done?”

Humpton also suggested that AI could help open up technical jobs to a wider pool of people. “People couldn’t come to work for Siemens because they didn’t have degrees in electrical engineering or mechanical engineering. AI is the first tool that actually makes the technology approachable,” she said.

“People can be more productive with the natural gifts they’ve been given without having to go on for an advanced degree,” she added. 

‘We’re going to have far fewer bank tellers’

New AI technologies are already changing how companies interact with their customers. 

“We’re seeing significant changes in terms of the numbers of employees that we need” due to AI, Timothy Wennes, chief executive officer of Santander US, said. “But what we are finding is that, even with the digitalization of many banking services and activities, there’s also a real important human connection. When there is a problem with a complex sale or service, that interaction is even more important.”

“We’re going to have far fewer bank tellers tomorrow, and at some point in time, you’re not going to need a bank teller,” he continued. “But…we’ve got people in our call centers, or people who are in our tech places, that are able to do more with generative AI.”

Panelists also discussed some of the challenges of building a skilled, AI-ready workforce.

“My problem now is that some people [have] become very comfortable with the current state of technology,” Alex Zhavonrokov, chief executive officer of Insilico Medicine, an AI drug discovery firm, said. “They even start coding using generative AI and rely very heavily on current transformer based methods.”

But current AI processes may not push the boundaries enough for what firms like Insilico want to do. “If you’ve got employees that are ready to go and really want to push the boundaries, you want to give them the ability to do that, and push them towards that.”

When it came to building an AI-capable workforce, Siemens decided it was more cost-effective to upskill its current workforce, rather than replace workers wholesale, Humpton said. 

Zemmel took a blunter view of companies trying to just hire their way to being AI-ready. “Companies have gone out and tried to hire all the Silicon Valley cool kids. That’s a good way to change the company dress code, but it doesn’t actually drive sustained business performance.” he said. 

This session was presented by McKinsey & Company. Discussion leaders included:

• Barbara Humpton, Chief Executive Officer, Siemens USA
• Timothy Wennes, Chief Executive Officer, Santander US
• Rodney Zemmel, Senior Partner, and Global Leader, Digital and Firmwide AI, McKinsey & Company
• Alex Zhavoronkov, Chief Executive Officer, Insilico Medicine
• Moderator: Andrew Nusca, Fortune