Meta’s enterprise AI plans go beyond agents

Meta’s enterprise AI strategy could extend well beyond the business agent it introduced in June, CEO Mark Zuckerberg told investors during the company’s second-quarter earnings call on Wednesday.

The initial agent is designed to help businesses with customer service, support, and other daily operations. But Zuckerberg described a broader portfolio that could eventually include APIs, business agents, compute sold directly to companies, and other services built for large customers.

“We see a large enterprise opportunity to sell to businesses, including APIs, business agents, potentially selling compute directly, and other services that we’re building for large customers.”

Meta’s first enterprise customers are its advertisers

Meta’s immediate focus is its existing base of advertisers. The company plans to offer AI agents that operate across its messaging apps and other services, allowing businesses to communicate with their customers through an AI interface.

Zuckerberg said Meta expects to be paid based on the results those systems deliver, mirroring the company’s advertising model. He positioned the effort as an extension of Meta’s existing relationships with many millions of advertisers and hundreds of millions of small businesses that use its platforms.

Recommended reading

Agentic software makes ERP migration the price of entry

“And, just like the ad system, effectively, we will get paid when we deliver results for those businesses.”

That approach could create revenue streams beyond advertising, which still drives most of Meta’s business, and subscriptions, which contribute a smaller share. Zuckerberg also acknowledged that selling to enterprise customers requires a “different muscle” from the one Meta has historically used.

Internal productivity tools could become products

Meta’s enterprise ambitions are not limited to small businesses. Zuckerberg said the company is increasingly developing coding, software development, and internal productivity tools for its own needs. Once those systems are in place, Meta believes it could offer them to both smaller and larger companies.

“We’re building coding and developing and internal productivity tools partially because we need to build them ourselves, and we need to make sure that we have tools that are tuned for ourselves.”

He added that Meta now sees an opportunity to serve customers outside its traditional advertiser base, although the company did not provide pricing, a product list, or a release schedule for those tools.

Compute sales compete with Meta’s long-term plans

Meta also discussed selling computing capacity to enterprise customers. The company said it currently has an opportunity to sell compute at “a significant premium” over what it paid for the hardware.

Zuckerberg warned that short-term revenue cannot come at the expense of Meta’s future AI infrastructure. He said it would be “foolish” to sell all of the company’s compute for immediate profit, describing the strategy instead as a portfolio combining short- and long-term infrastructure plans.

“As we get closer to personal superintelligence, we are . . . going to need hardware that allows you to seamlessly interact with it.”

Agents are also aimed at consumers

The earnings call covered ambitions beyond enterprise software. Meta is promising consumers personal AI agents, along with AI-powered smartglasses that can interact with the world in front of the wearer.

The company is also using large language models to build and expand its social-app portfolio more quickly. Recent launches include an app for Marketplace sellers, another for Facebook Groups, and one for vibe-coded games, along with other experiments.

Zuckerberg said Meta plans to develop more app ideas and use its recommendation systems to distribute them to people likely to find them interesting. He also expects it to become easier for the company to ship new apps, but did not specify which products are next or when they will arrive.

“I expect it to become a lot easier to ship new apps. So we are planning to build out more ideas and use our recommendation systems to scale them to the people who will find them interesting.”