Zuckerberg's 24/7 agent promise runs into Meta's trust problem

Zuckerberg's 24/7 agent pitch depends on Meta fixing a problem it has never solved: consumer trust.

On the Q2 2026 earnings call, he described agents that work around the clock on their owners' behalf, across health, relationships, finances and more: "Soon we will have agents that can work 24/7 on your behalf to help you achieve your goals and improve your life, your health, your relationships, your finances, whatever you want." The 24/7 promise is not unique to Meta. One-click cloud agents like Kimi Claw already run around the clock with no local setup.

He drew a hard line between where agents have caught on and what mass adoption will require. "The first domain that agents have really taken off in is coding. But engineers are more technical and willing to spend time making those agents work, so to build great personal agents, this needs to be a great consumer product that just works out of the box and is easy enough for billions of people to adopt and use." The coding-first pattern shows up in benchmark data too: agents get far on coding tasks but stall where enterprises need them, according to a 957,253-record review of 30 benchmarks.

That reads as a deliberate contrast with where the biggest AI labs put their energy, as The Verge's report on the call notes. Anthropic and OpenAI have invested heavily in agents aimed at coding and enterprise work. Google pushed into personal agents with its Gemini Spark tool, which The Verge found very good "to an almost creepy degree," and has since added Mac remote control and third-party app integrations, per our update. Meta is betting on the opposite path: consumer-grade first, engineers later.

"Just works out of the box" is a different engineering problem from the one that produced the coding wave. Developers will put up with a finicky agent and tolerate rough edges for the payoff. The billions of users Zuckerberg is targeting will not. Meta has said little about how it plans to cross that gap, and the gap is not unique to consumers: agents that work in demos often stall on real production work, as our look at AI agents in production shows.

Who is positioned where

The call laid out a competitive field with very different starting points:

Player	Where things stand
Meta	Previewing consumer agents for health, relationships and finances; more than 1 million businesses use its WhatsApp and Messenger agents every week.
Anthropic / OpenAI	Heavy investment in agents for coding and enterprise work.
Google	A personal agent push via Gemini Spark; ecosystem access to users' email and documents.
Microsoft	Ecosystem access to users' email and documents.

What Meta actually brings

Zuckerberg called personal agents "the foundation for our next wave of products and revenue lines in the months and years ahead." That is a revenue claim, not a research statement. The closest existing product Meta can point to is its business agents on WhatsApp and Messenger, used by more than 1 million businesses every week and now rolling out to Instagram. Instagram itself has passed 2 billion daily active users, Zuckerberg said.

Underlying the product push is model work. Muse Spark, Meta's AI model launched last quarter, received a 1.1 update with stronger coding capabilities, the exact domain Zuckerberg acknowledged led the agent wave.

The gap that capital can't close

The bill for this bet is enormous. Meta expects $130 billion to $145 billion in capital expenditures in 2026 and is building a 1 gigawatt data center campus with BlackRock. Restructuring reportedly moved 7,000 staffers onto new AI initiatives, alongside about 8,000 layoffs in May.

None of that touches the structural problems The Verge's report flags. Meta's AI efforts still lag the bigger labs; Anthropic, for one, already has a 30,000-engineer enterprise distribution channel through Cognizant, per our coverage. Unlike Google or Microsoft, Meta has no ecosystem access to people's email or documents. And it carries a trust deficit: the report points to a "different creep problem" with its smart glasses and to general distrust of Meta as reasons users might keep its agents away from their data.

The "creepy" finding hurts the whole category. The Verge's reviewer found Google's own personal agent effective and unsettling at the same time. That is the bar Zuckerberg is racing, and reviewers have already used the same words about Meta's own glasses.

Zuckerberg said Meta will share more "soon." That is a timeline, not a reason. Engineers, as he noted, will tolerate friction and bad setup. Billions of people need a reason to hand an agent their finances in the first place, and he has not said what that reason is.