After Zoom went public in 2019, Yuan had scaled back to a five-day workweek, but now he’s back to putting in extra time to help steer Zoom toward what he sees as its future: harnessing AI to boost productivity and, ultimately, enable people to work less—himself included. Despite the increased workload for now, Yuan insists he’s no workaholic; he’s simply investing the time to ensure a more efficient, AI-powered future for all.

Of course, he is also investing that time to ensure Zoom’s success in its AI efforts, which aim to demonstrate that the company is no longer just the videoconferencing startup that took the pandemic world by storm in 2020, but struggled to maintain momentum after offices reopened. Zoom stock, after all, plummeted to near-IPO levels in 2022 and hasn’t budged much since. 

“We are way more than just meeting conferencing,” Yuan told Fortune in an exclusive interview in advance of the company’s Zoomtopia conference, which today unveiled a slew of new AI work and business features. These include upgrades to its Copilot-like “AI Companion” that allow users, at no extra cost, to generate content and search across Zoom’s workplace platform and connect data from third-party apps including Microsoft Office and Google Docs, as well as ServiceNow, Glean, Workday, Box, Asana, and HubSpot. 

Though he agreed the landscape of AI workplace tools is crowded, he insisted that Zoom’s AI moves and primary focus on the workplace position the company to challenge Microsoft 365 and Google Workspace for AI dominance at the office. 

“There’s too many [other] things on their plates,” Yuan said of his competitors, touting Zoom’s speed of innovation and culture of customer care.  

Some analysts are doubtful

But some analysts are doubtful Zoom can take on Microsoft’s and Google’s massive resources and long history in work solutions. Daniel Newman, CEO and chief analyst at the Futurum Group, told Fortune that while Zoom’s platform and AI features may win when it comes to its user-friendly experience and service, “these are also companies that can build their own models, they can deploy data center infrastructure at incredible scale, and they own the office environment—processing email, spreadsheets.” 

Others are prepared to give Zoom the benefit of the doubt, as it has shown it can focus on collaboration and can form relationships with makers of other Microsoft and Google competitors, including Salesforce and Slack, said Ross Rubin, founder and principal analyst at Reticle Research. “Every company is feeling compelled to roll AI into their [products]; the next challenge will be how they work together to get out of their silos,” he said. 

Dan Ives, managing director at Wedbush Securities, added that Zoom has a big opportunity in AI work solutions because it is already installed and used on so many devices. “They’re in a very strong position to monetize,” he said. “Over the coming years, they have to prove it out—but this is a big enough ocean for more than two boats.” 

Kena Betancur—Getty Images

Contrasting approaches to AI models

One of the biggest differences between Zoom’s approach to AI and that of Microsoft and Google is access to the most sophisticated frontier models. Both of the hyperscalers have their own powerful AI models to turn to—Google’s Gemini and Microsoft’s partnership with OpenAI—as well as their own cloud and data center infrastructure. 

Yuan says Zoom’s approach is different in that it uses a variety of models, including its own LLM and those from OpenAI, Anthropic (Zoom is an investor), and Meta. “Essentially, we have four models together to compete [with] any of the single models,” he said, adding that this “federated” AI approach allows the company to choose the best model for each feature. For example, Zoom’s AI-generated meeting summaries are “much more accurate than any of our competitors,” he said. 

But Newman of the Futurum Group said using multiple models will quickly become “table stakes.” After all, Microsoft, Google, Amazon, and others all have the ability to use open-source models, like Meta’s Llama or those from Mistral, or licensed closed models like OpenAI’s ChatGPT. “That’s a design choice,” he said. 

Ives agreed that Zoom’s lack of its own frontier model could put it at a disadvantage: “That will be one of the uphill battles that they have to get around,” he said. 

Zoom has faced other AI challenges

Zoom has faced other AI challenges over the past two years: There was blowback in 2023 over a change in the company’s terms of service that was widely interpreted to mean Zoom was training AI models on customer data, but it quickly published a blog post asserting that it “does not use any of your audio, video, chat, screen sharing, attachments, or other communications like customer content (such as poll results, whiteboard, and reactions) to train Zoom’s or third-party artificial intelligence models.” 

Yuan claimed Zoom was the first vendor to publicly make that commitment, but adds that not all users read or understand the terms of service, and that the company “learned a lesson” about improving communication. Zoom does train Zoom AI on meetings data, but only internal company meetings he said, and even that permission is not turned on by default. Zoom also partners with third-party vendors to gather data, he added, and buys some data as well. 

Battling deepfakes is also a challenge for Zoom—one that has made media headlines, like when a deepfake caller posing as a Ukrainian official showed up on a Zoom call with Sen. Benjamin Cardin, chairman of the U.S. Foreign Relations Committee. But Yuan said the company is working on a feature to detect deepfakes: “We have high confidence that the problem will be fixed. Hopefully very soon you will see some announcements,” he said. “Authentication is key.” 

Jobs are not at risk with an AI-driven workplace, CEO says

Like other tech CEOs touting their workplace AI solutions, Yuan maintains he does not believe jobs are at risk—even though more productivity may mean companies need fewer workers. Instead, he says, employees will simply be able to do more in less time, and potentially work four days per week instead of five. Fears of widespread job loss did not pan out in the internet or smartphone eras, he noted. “As we further improve productivity, there is more time to do more meaningful things,” he explained, adding that in the future, multiple digital agents will take on many daily tasks. 

Yuan did not shy away from previous comments he has made about the potential for users to ultimately train a digital assistant on their own data and have a digital twin of themselves show up at their work meetings on Zoom.

“Imagine Company A and Company B are working together to negotiate a contract and they send their lawyers’ digital twins to come up with a preliminary draft [in] 20 to 30 minutes,” he said. “And afterward the two CEOs look at the draft without any problem. That’s the future of work.” 

Zoom may offer the better experience, said the Futurum Group’s Newman, but that experience coexists with powerful alternatives. “Microsoft is not going to just lay down,” he explained. “Big companies have too many moats with how they can apply AI across the portfolio, and I think it’s going to actually create a tougher, more challenging climate.” 

That doesn’t mean Zoom is out of luck, he added. “It just means I think they’re going to have to fight harder to win deals, despite whether they have some advantages with individual features.”