Meta has created a dedicated enterprise AI business that will package agents, developer tools, and infrastructure for companies, moving the owner of Facebook, Instagram, and WhatsApp more directly into a corporate software market where its commercial position has historically centred on advertising rather than internal business systems.

Meta Enterprise Platform will initially bring together the company’s Muse agent, Meta Business Agent, Muse API, Muse Code, and other parts of its AI stack. Chirantan “CJ” Desai, who has led MongoDB, will join Meta as chief enterprise platform officer and report directly to Mark Zuckerberg.

The announcement establishes the organisational structure for the push but leaves much of the product proposition unresolved. Meta has not published pricing, a detailed availability timetable, European deployment arrangements, service commitments, administration tooling, or a full explanation of how the products will operate together inside an enterprise environment.

That makes the launch less a finished software-suite release than a declaration that Meta intends to compete for a larger share of corporate AI spending. Microsoft, Google, Salesforce, ServiceNow, SAP, Oracle, and specialist AI providers have spent the past several years embedding models and agents into existing enterprise products, while Meta has largely supplied models, advertising tools, and consumer-facing services around the edges of that market.

Enterprise software requires a different relationship

Meta already has extensive commercial relationships with businesses, but most are built around companies reaching customers through advertising, commerce, messaging, and social platforms. Enterprise technology procurement asks a different set of questions because software is being trusted with internal data, employee identities, regulated information, and operating processes rather than primarily customer acquisition.

A large organisation evaluating an agent platform will want to know where its information is processed, how administrators can limit permissions, which actions are recorded, how employees authenticate, what happens when a system fails, and whether the supplier can meet contractual requirements around uptime, security, retention, and support.

European deployments add further requirements around data protection, sector regulation, employee consultation in some markets, and the EU AI Act where particular systems fall within its scope. None of those issues prevents Meta from entering enterprise software, but they mean consumer-scale adoption does not automatically translate into corporate adoption.

Muse gives the company a potentially important starting point because it is designed around taking actions rather than simply responding to prompts. Meta introduced Muse earlier this month as a personal agent running inside a dedicated secure virtual machine, while Meta Business Agent was launched separately for customer-facing business interactions.

Corporate deployment exposes the security problem more sharply. An assistant acting for one consumer can be isolated relatively tightly, whereas a business agent may require controlled access to shared systems used by hundreds or thousands of employees. Identity, permissions, audit trails, approval steps, and separation between customer environments therefore become central product features rather than background infrastructure.

Meta assembles another AI route to market

The enterprise move broadens a strategy that has not followed one consistent commercial model. Meta has released open-weight models, developed consumer assistants, embedded AI across its advertising products, and invested heavily in computing infrastructure, while retaining advertising as the engine that funds most of that spending.

Its earlier work on open-weight agents capable of running on local hardware pointed towards one route into organisations that want greater control over deployment. Meta Enterprise Platform creates another, in which the company itself becomes a supplier of AI products and services to corporate customers.

The initial announcement names Muse and the company’s developer tooling but does not describe Llama as the organising layer of the new platform. Enterprise buyers will therefore need to understand whether Meta intends to compete mainly through models, through complete applications, through infrastructure services, or through a combination of all three.

Desai’s appointment suggests that the company recognises the difference between developing AI technology and selling enterprise platforms. Meta says his background includes senior roles at MongoDB, Cloudflare, and ServiceNow, giving the new division leadership drawn from the procurement, support, security, and ecosystem requirements surrounding large corporate software deployments.

The competitive pressure extends beyond software licences. AI companies are spending heavily on data centres, processors, networking, and energy, and those assets become easier to justify when they support several revenue streams. Enterprise AI offers Meta another route for monetising infrastructure already serving consumer products, advertising systems, and research.

The corporate market can nevertheless be slower and less forgiving than the consumer businesses in which Meta built its scale. Large organisations tend to adopt new platforms alongside existing systems rather than replace them immediately, while regulated companies may spend months evaluating security architecture and contractual terms before production deployment begins.

Meta therefore enters with considerable computing capacity and a rapidly expanding AI portfolio but without the decades of installed enterprise workflows owned by incumbent software suppliers. Its first task is not simply to show that its agents can perform useful work; it is to demonstrate that companies can govern them, integrate them, procure them, and operate them with the controls expected of business-critical software.