Meta enterprise AI plan stretches from agents to compute sales
Mark Zuckerberg told investors Meta sees business AI revenue in agents, APIs, compute and internal tools, widening its push beyond ads.
By Jordan Bell · Startups & Deals Reporter
· 3 min read
Meta enterprise AI is becoming a broader revenue target for the company, CEO Mark Zuckerberg told investors on Wednesday’s second-quarter earnings call. For everyday investors, the pitch matters because Meta is describing AI as more than a feature inside Facebook, Instagram and WhatsApp: it is also a possible business line alongside advertising and subscriptions.
Zuckerberg said Meta sees a sizable chance to sell AI-related products to companies, including application programming interfaces, business agents, computing capacity and services built for larger customers. An application programming interface, or API, lets outside software connect to a company’s technology. Compute means the data-center processing power needed to train or run AI systems.
Meta’s core business still depends heavily on ads, with subscriptions playing a smaller role. The enterprise push suggests the company is looking for additional ways to turn its AI spending into revenue, though Zuckerberg also acknowledged that selling to big companies requires skills Meta has not historically relied on as much. He called enterprise sales a “different muscle” for the company.
What is Meta’s enterprise AI plan?
Meta’s first focus is businesses already using its platforms, especially advertisers and small companies. Zuckerberg said Meta plans to offer AI agents that can communicate with customers through messaging apps and other surfaces, with payment tied to business outcomes.
AI agents are software systems designed to take actions for a person or company, rather than only answer questions. In Meta’s case, Zuckerberg described them as tools that could help businesses interact with customers through an AI interface.
Reuters reported in June that Meta had launched an enterprise-focused AI business agent meant to automate tasks such as customer service, support and daily operations. On the earnings call, Zuckerberg framed that as only one part of the opportunity, saying Meta can build on relationships with millions of advertisers and hundreds of millions of small businesses that use its apps.
“Just like the ad system,” Zuckerberg said, Meta expects to be paid when it produces results for businesses. That model would make the AI products more familiar to advertisers already used to spending on Meta based on performance.
Internal tools could become products
Zuckerberg also pointed to another path: selling tools Meta first builds for itself. He said the company is developing coding, software development and internal productivity tools partly because Meta needs them for its own employees.
Once those tools exist, Zuckerberg said Meta sees a chance to offer them to small businesses and larger companies. The company did not give pricing, timing or product names for those potential enterprise tools on the call.
Meta is also weighing whether to sell computing capacity directly to enterprise customers. The company said more than once on the call that it currently has an opportunity to sell compute at a significant premium to its own cost.
Zuckerberg warned against treating that as a short-term cash grab. He said it would be “foolish” to sell all of Meta’s compute for near-term profit, because the company still needs infrastructure for its own AI goals. He described the plan as a portfolio that balances shorter-term revenue with longer-term needs.
The call also covered Meta’s consumer AI plans. Zuckerberg discussed personal AI agents, AI smart glasses that can respond to what a wearer sees, and the company’s use of large language models to build social apps faster. Large language models are AI systems trained on large amounts of text and other data to generate or interpret language.
Recent Meta experiments cited around that effort include apps for Marketplace sellers, Facebook Groups, games and other tests. Zuckerberg said he expects it to become easier for Meta to launch new apps and use its recommendation systems to find users who may be interested in them.
This story draws on original reporting from TechCrunch.
