Meta taps MongoDB's ex-CEO to lead enterprise AI push, but Llama's role stays murky

Meta is now positioning enterprise AI as a major new business pillar alongside advertising and consumer products. The company has named CJ Desai, who until this week led MongoDB, as Chief Enterprise Platform Officer to oversee the initiative, which Meta is calling Meta Enterprise Platform.

Desai departed MongoDB effective immediately—less than a year into his tenure as CEO—with former CEO Dev Ittycheria returning as interim leader. He will report directly to Mark Zuckerberg at Meta and inherit a mandate to package the company's AI models and agents for deployment within corporate environments.

What Meta is offering developers and businesses#

Meta plans to surface its full technology stack to enterprises, beginning with a suite of agent and coding tools: the Muse agent, Meta Business Agent (which launched in June to handle customer interactions), Muse API, and Muse Code. The company introduced Muse earlier this month as a consumer-facing personal assistant. Both Muse API and Muse Code are positioned as developer-facing products that pit Meta against OpenAI, Anthropic, and Google in the race for engineering teams building agents and coding workflows.

Muse is already moving beyond Meta's ecosystem—Shopify has begun integrating it across its stores, though Amazon chose to block it.

The missing enterprise details#

Meta released no enterprise pricing, general availability dates, or service-level terms for its Muse products on launch day, despite developers already having access to some tools. Muse Code entered beta in August, and Meta began accepting payments for its Muse Spark model through the API starting in July at $1.25 per million input tokens and $4.25 per million output tokens.

Developers scrutinizing these offerings will need clear answers on how Meta handles model updates—a critical concern given how easily a model change can break production systems. Enterprise buyers will also be looking for specifics on data retention, customer data usage, tenant isolation, identity controls, and compliance certifications. While Meta has published a security and safety approach for Muse, these details remain absent.

Llama's uncertain future in the enterprise strategy#

Llama, the open-weights model family Meta has promoted for years to developers wanting to run and fine-tune models on their own infrastructure, conspicuously absent from Monday's enterprise announcement. Meta has not clarified whether Llama will factor into the Enterprise Platform at all.