Meta is building an AI version of Mark Zuckerberg so employees can ‘talk to the boss’

At a glance:

• Meta’s Superintelligence Labs is creating a photorealistic AI avatar of Mark Zuckerberg for internal use.
• The digital twin is trained on Zuckerberg’s mannerisms, tone and strategic thinking to let staff "talk to the boss".
• The project is separate from a separate "CEO agent" that assists Zuckerberg directly and targets Meta’s roughly 79,000 employees.

Meta’s new AI avatar of its founder

Meta is developing a lifelike, AI‑powered digital version of Mark Zuckerberg that can hold real‑time conversations with employees. The effort, described by four people familiar with the work, is being led by Meta’s Superintelligence Labs. Training data include publicly available statements, internal strategy documents, and recordings of Zuckerberg’s speech patterns so the avatar mimics his mannerisms and tone. Zuckerberg himself is hands‑on, reviewing the model and participating in testing sessions.

How the avatar differs from the "CEO agent"

The avatar is intended to stand in for Zuckerberg in employee interactions, whereas a previously reported "CEO agent" is a tool that helps the real Zuckerberg retrieve information faster. The two projects are distinct: the avatar is outward‑facing, designed to give staff a sense of connection to the founder, while the agent is an internal assistant that streamlines his own workflow. Both underscore Meta’s broader push to embed generative AI across the company’s operations.

Technical challenges and scale

Creating a photorealistic conversational agent is computationally intensive. The model must render realistic facial movements and respond without perceptible latency, which demands massive GPU clusters and sophisticated inference pipelines. Meta’s engineers are reportedly grappling with the balance between visual fidelity and real‑time interaction speed, a problem that has limited similar attempts elsewhere.

Zuckerberg’s hands‑on AI involvement

According to insiders, Zuckerberg spends five to ten hours each week writing code and attending technical reviews for various AI projects—a rare level of direct engagement for a CEO of a $1.6 trillion company. He has publicly championed a “personal superintelligence” vision, positioning Meta to close the gap with rivals like OpenAI and Google. On a January earnings call, he said AI‑native tooling would “elevate individual contributors and flatten teams,” signaling that the avatar is part of a larger cultural shift.

Meta’s previous forays into AI characters

Meta’s experience with AI‑driven personas dates back to September 2023, when it launched celebrity‑based chatbots featuring Snoop Dogg, Tom Brady, Kendall Jenner and Naomi Osaka. Those bots were discontinued in summer 2024 after low user adoption. An AI Studio later let creators build custom characters, but it sparked controversy over sexually explicit personas, prompting Meta to restrict teenage access in January 2024. The current avatar project builds on those lessons, aiming for an internal, controlled use case rather than a public product.

Industry context and competitive landscape

Meta is not alone in experimenting with leadership clones. Uber’s CEO Dara Khosrowshahi confirmed his team built an AI version of him earlier this year. However, Meta’s initiative is larger in scale, targeting a workforce of roughly 79,000 employees and serving a strategic purpose: fostering a sense of direct connection to a founder who is otherwise hard to reach. The move reflects a growing trend among tech giants to use synthetic avatars for internal communication, knowledge sharing and brand reinforcement.

What to watch next

The avatar remains in an early development stage, and Meta has not disclosed a rollout timeline. Observers will be watching for performance benchmarks, employee adoption rates, and any policy guidance around synthetic representations of real people. If successful, the project could set a precedent for other large enterprises seeking to humanise AI‑driven internal tools.