AI avatars are crashing the C-suite—is this a scalable fix for exec overload, or a shortcut that chips away at trust?
If you tuned into Klarna’s Q1 2025 earnings call in mid-May — curious, perhaps, about the BNPL giant’s IPO-launch status — you encountered a stoic introduction by CEO and Co-Founder Sebstian Siemiatkowski, donning a clay-toned jacket and dark brown button-down, and standing before a blurred, warm interior seemingly out of an IKEA catalogue.
Only this wasn’t Siemiatkowski: It was an AI avatar made in his image, delivering a scripted, and somewhat stilted, 80-second speech. With video-audio alignment resembling a B+ lip sync and the avatar blinking his eyes the uncanny way Disneyland’s animatronic characters might bat their eyelids, the automated intro wasn’t exactly charismatic.
But, presumably, charisma wasn’t the point. The deployment of an AI avatar generated significant buzz around Klarna’s earnings call — almost guaranteeing that this article is the umpteenth time you’ve heard about this stunt; the ploy may stay longer in your memory than the earnings figures Siemiatkowski’s cyber-doppelgänger shared.
Despite the backlash it’s received from customers and employees for its aggressive push into automation, including overzealously using AI in sensitive settings like customer service (and now backtracking on those efforts), Klarna, via video avatar, doubled down on its AI focus and highlighted, in its view, the boon that its AI ambitions represent for investors. AI, Klarna says, enabled the company to achieve profitability for a fourth quarter in a row, and boost revenue per employee to almost $1 million.
Leadership at another major tech company followed a similar earnings-call playbook later that week. Zoom CEO Eric Yuan used an AI avatar for his opening comments. The avatar cited its existence as “just one example of how Zoom is pushing the boundaries of communication and collaboration.” Built using Zoom’s Clips tool, the digital version of Yuan functions both as marketing for a specific product, as well as a reflection of Zoom’s larger AI goals.
By leaning into avatars with Clips, Zoom is entering a nascent (and arguably unproven and controversial) space that already features several prominent players. London-based Synthesia recently passed $100 million in ARR, and has secured over $330 million in venture investment from the likes of Kleiner Perkins and Accel. (Zoom and Synthesia list each other as clients; Zoom did not respond by press time whether its Clips avatar feature is a whitelabeled version of Synthesia’s product.)
Accel has also invested in SF-based YC spinoff Tavus, which aims to “bring a human dimension to AI agents” through video chat interfaces with avatars; it most recently raised an $18 million Series A in August 2023. Other startups abound, including D-ID, Colossyan, and Elai.
According to Mark Loehr, CEO of virtual meetings management solution OpenExchange, and who works closely with Zoom and Yuan, executives who are short on time are willing to try tools like avatars for pre-recorded components of earnings calls given the potential time-saving benefits. Loehr said he’s already made five pre-recorded avatar videos for internal weekly messages to his team.
“I don’t have to go spend half an hour doing a five-minute recording,” he said. “I just type it out and it’s really good to go.”
Loehr sees additional promise in the technology as a way to deliver the same message in several languages, potentially simultaneously.
To Yuan of Zoom, these avatars represent the tip of the iceberg. Last year, he told The Verge he envisions letting his AI avatar attend meetings on his behalf, and, with the right data uploaded about Yuan, even make decisions on his behalf. Yuan said at the time the technology is “not there yet.” Zoom declined to share with AI Nexus how Yuan would determine if or when such technology was appropriate for public use.
In their current form, the avatars are fed pre-written scripts; models then transform the words into audio tracks in the speaker’s voice, superimposed on a video-generated humanoid. Though creepy, there’s relatively little room for hallucination or deviation from the planned path. But these tools carry risks, especially if their complexity and convincingness improve over time.
For one, employees and investors may be offended by the fact that an executive would rather automate their interactions with core constituents than take the time to interface with them organically. It’s especially hard to justify outsourcing executive work to the digital realm if these same executives frequently tout the benefits of “return to work” policies and other strategies centered around in-person congregation.
AI avatars also carry liability and authenticity concerns. As Otter CEO Sam Liang put it, “Anything the avatar CEO says, the real CEO should take responsibility for.” That may result in CEOs playing statement Whac-A-Mole, catching up retroactively with AI-generated commitments and strategic choices. Furthermore, executives known for using avatars — even uncanny ones — may see other entities make counterfeit versions of the executive’s messages, leading to liability issues as well as a loss of narrative control. (The line between “generative video solutions” and “deepfakes” is razor-thin, after all.)
For their part, software companies like Synthesia claim to uphold stringent content-moderation policies, including safeguards that prevent non-consensual use of a person’s likeness. Synthesia uses a laptop’s webcam for avatar creation and relies on red team tests to identify lapses in these systems. Additionally, Alexandru Voica, Head of Corporate Affairs at Synthesia, said via email that video flagged by internal moderation systems is passed onto human moderators for review. It includes more than 60 content moderation policies aiming to mitigate the harms of video generation. But Synthesia is just one of many generative video tools — some of which may find lax moderation policies a lucrative market wedge, regardless of the social implications.
Some of these concerns may seem irrelevant to leaders looking to use AI avatars in discrete settings. But the use of these tools may signal to investors and employees what a leader’s risk tolerance is. That may serve a tech-forward company like Klarna somewhat well, but not, say, a medical company’s leadership. And, if avatars’ emotionlessness remains a feature, not a bug, then their injection into workaday office rituals will only re-entrench a lack of collective engagement with, or investment in, corporate life: Lifeless matter that speaks but fails to enchant.
