Meta CEO Mark Zuckerberg says many AI agents are still too hard to use for everyday people, even as progress accelerates.
AI agents may be getting smarter, but Meta CEO Mark Zuckerberg says most are not yet ready for real-world use.
On Meta’s first-quarter earnings call. Zuckerberg argued that only a small share of current agents meet what he called an informal “mother test” meaning a product you could confidently recommend for someone who does not want technical hurdles.. His point was blunt: while there are many agents showing promise. he would not yet want to hand most of them to his mother.
This highlights a key divide in the AI market: capability versus usability. An agent can look impressive in demos yet still fail the day-to-day experience that businesses and consumers actually need.
Zuckerberg singled out the setup required for some agents. describing a process that involves installing software locally. accessing a computer’s terminal. and configuring systems before the agent can be used.. He said that only a “small numbers of millions of people” are going through that kind of setup today. even as Meta is aiming for broader reach.
In this context, Meta’s focus is less about chasing a narrow deadline and more about polishing the experience until “it just works.” Zuckerberg framed ease of use as a gating factor for whether agents move from experimental tools to practical products.
For the industry, this matters because mass adoption typically depends on frictionless deployment. If agents require specialized setup, their value stays limited to enthusiasts and technical users, slowing the shift from curiosity to routine workflows.
The call also touched on coding agents, an area where multiple companies have been investing heavily.. Zuckerberg said Meta is not necessarily a developer tools company, though he did not rule out building coding-oriented capabilities.. He emphasized that people sometimes overstate coding as the whole path to “self-improvement. ” arguing that coding is only one ingredient rather than the entire solution.
In other words, Meta’s messaging suggests it views coding as part of a bigger agent strategy rather than the end goal itself. That stance could shape how the company allocates engineering effort as the market races to commercialize AI agents.
Ultimately, the “mother test” framing is a reminder that AI’s next phase depends on product design as much as model performance. When agents become easy enough for everyday users, they stop being a novelty and start turning into economic tools.