Meta hires MongoDB CEO CJ Desai to lead new enterprise AI business
SiliconANGLE Maria Deutscher ● Covered by 4 sources
Meta is launching an enterprise AI unit and put MongoDB CEO CJ Desai in charge. It’s a sign Meta wants more than consumer bots — and maybe a slice of business spending too.
Based on reporting by SiliconANGLE, Maria Deutscher — 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 building a new enterprise AI business and handing it to CJ Desai, the former CEO of MongoDB. The new unit is called Meta Enterprise Platform, and Desai will serve as chief enterprise platform officer. MongoDB’s shares dropped more than 18% on the news, and the database company brought back Dev Ittycheria as interim CEO.
Meta has said surprisingly little about the business itself, but the outline is clear enough. The first push will be to grow adoption of Muse among enterprises and developers. The company is also using the launch to sell Muse Code and Meta Business Agent, two products that point in very different directions: one for software building, the other for customer support.
Muse Code arrived in August, and Meta says its fan-out system is a key difference from the consumer version of Muse. Instead of tackling a complex coding job in one pass, it breaks the work across multiple subagents, which the company says makes the process much faster. Meta Business Agent is the simpler pitch: a chatbot for WhatsApp that can answer product questions, offer buying suggestions and handle some technical problems.
The enterprise platform also includes Meta’s Muse API, which gives developers access to the large language models behind these agents. That matters because Meta is not just selling an app here; it is trying to turn the models themselves into a business product. Muse Spark 1.3, the company’s newest model, launched earlier this month and beat GPT-5.6 Sol on AutomationBench, while also doing better on five coding benchmarks.
Meta says it trained Muse Spark 1.3 with cost in mind. The model uses about 25% fewer tokens than its predecessor on coding tasks, helped in part by making less use of external tools. Security and privacy are part of the sales pitch too, with Desai saying they are built in from the start. Meta plans a feature called Muse Confidential VM later this year, which will isolate instances so tightly that even Meta can’t access their contents.
That is the real tell here. Meta is trying to make enterprise AI feel safer, cheaper and less embarrassing than the usual chatbot demo, which is exactly what businesses will ask for before they hand over real work.
My take — AI-written commentary, not fact-checked reporting
This is Meta doing what Meta does best: turning a model story into a distribution story. The interesting bit isn’t the launch fluff; it’s the push toward security, privacy and enterprise controls, because that’s where the actual money hides. Open models win when companies want leverage, and closed systems win when they want someone else to carry the risk — preferably in a neat little VM.
Read more about this at: SiliconANGLE
