Meta taps CJ Desai to lead new enterprise AI unit

Meta's move puts it more directly into the business of selling AI to companies, rather than only serving consumers and advertisers. The company has said the initial goal is to broaden adoption of its Muse agent among enterprises and developers, with Muse Code and Meta Business Agent — a chatbot that fields customer inquiries via WhatsApp — rounding out the go-to-market push. Underneath those products sits the Muse API, which gives developers access to the large language models powering the agents.

Desai arrives with a mandate that is still thin on public detail. Meta debuted its newest LLM, Muse Spark 1.3, earlier this month, saying it outperformed GPT-5.6 Sol on AutomationBench and scored higher on five coding benchmarks, while using about 25% fewer tokens than its predecessor on coding tasks. On the security side, Meta plans a feature called Muse Confidential VM that isolates Muse instances so that not even Meta can access their contents, plus a "continuous audit" so users can verify it works.

The stakes appear to hang on whether that security story lands with cautious corporate buyers. Meta's incentive is to make Muse more appealing to risk-averse enterprises worried about AI hallucinations, and one path would be integrations with third-party cybersecurity tools — the kind of connectors Anthropic offers for sending Claude usage logs to external breach detection platforms. It is also believed Meta's enterprise portfolio will include infrastructure services, following July reports from Bloomberg and CNBC that it intends to sell some excess data center capacity. How much of this materializes, and how quickly, looks likely to determine whether the new unit becomes a real enterprise contender or a side bet.