Meta hired MongoDB’s CEO to build its enterprise AI business — but Llama is missing
The New Stack Amanda Caswell ● Covered by 4 sources
Meta hired MongoDB’s CEO to run a new enterprise AI push. Big question: does this mean less Llama and more locked-down Meta tools?
Based on reporting by The New Stack, Amanda Caswell — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Meta is taking a swing at enterprise AI, and it picked a notable operator to do it. On Monday, the company said it is building Meta Enterprise Platform and hired MongoDB CEO CJ Desai to run it. Desai will report directly to Mark Zuckerberg, who called the effort the company’s “next major pillar,” alongside ads and consumer apps.
The pitch is straightforward enough: take technology Meta already built for consumers and advertisers, then sell it to businesses and developers that want to run it inside their own operations. Zuckerberg said the stack will start with Muse agent, Meta Business Agent, Muse API, and Muse Code. That puts Meta closer to the same customers OpenAI, Anthropic, and Google are chasing on engineering teams building agents and coding tools.
But the launch leaves out a lot of the details buyers usually want first. Meta did not give enterprise pricing, general availability dates, or service terms for Muse API or Muse Code. Developers can already use both, though. Muse Code has been in beta since August, and Meta started charging for Muse Spark through its API in July, at $1.25 per million input tokens and $4.25 per million output tokens.
The bigger omission is Llama. The model family Meta spent years pitching as its open-weights answer for teams that wanted to run and fine-tune models on their own infrastructure does not appear in the new enterprise description, and Meta has not said whether it will be part of the platform. That matters because Meta has also been mixing signals: it has released open weights for the smaller Muse Glimmer model, and it has promised open weights for Muse Spark, while now building a more controlled enterprise offering around Muse.
Desai’s background fits the problem Meta is trying to solve. At MongoDB, he had already been pushing on persistent memory, automated embeddings, and other pieces meant to help agents survive in production. That’s the real issue here. Businesses do not just need a model; they need memory, current data, identity controls, and a way for agents to act inside the systems employees already use. Meta says security and privacy are built in from the outset, but it did not address data retention, training on customer data, tenant isolation, or compliance certifications on Monday. Those answers will matter a lot more than the slogan.
Meta does have one advantage: reach. Zuckerberg points to billions of people using Meta products and hundreds of millions of businesses on its platforms, many of them already relying on Facebook, Instagram, and WhatsApp for marketing and customer service. The hard part is getting larger companies to add yet another platform when their teams have already spent years building around OpenAI, Anthropic, Google, and others. Meta’s new business is real. Whether it’s a platform or just a bundle of promising parts is still up for debate.
My take — AI-written commentary, not fact-checked reporting
Meta keeps trying to turn distribution into destiny, which is very on-brand and only sometimes works. Enterprise buyers won’t hand over internal data because Zuckerberg called something a “next major pillar”; they’ll want terms, controls, and a clean story about Llama. Right now this looks less like a finished platform and more like Meta asking the market to trust the vibe, which is not exactly an enterprise sales strategy.
Read more about this at: The New Stack