Meta began rolling out autonomous task execution capabilities to Meta AI on July 24, 2026, giving its consumer assistant the ability to connect to calendars, deliver scheduled daily briefings, run multi-step research workflows, and generate slide decks without re-prompting. The update, announced by Meta, moves the assistant from answering questions to taking independent action on behalf of users.
What Ships
According to Meta’s announcement, Meta AI can now generate a recurring daily briefing that pulls from a connected calendar, flags scheduling conflicts like double-bookings, and delivers a summary at a user-defined time. Users set up a task once, and the assistant handles it on schedule: weekly meal plans, product restock alerts, or afternoon trend updates all run without additional prompts.
The assistant can also research topics by synthesizing information from the open web and from creators and communities across Meta’s apps, then convert the results into reports, presentations, or slides. Users can steer output mid-generation, redirecting focus, adjusting tone, or cutting sections while the assistant is still working. All generated content collects in a single workspace within the app, as Unite.AI reported.
Meta’s own demonstrations lean domestic: scouting Facebook Marketplace for kitchen renovation furniture within a budget, building a week-by-week half-marathon training schedule, or selecting a restaurant and checking calendar availability for a birthday dinner. These are capability previews. How reliably the assistant executes across real-world calendars and connected accounts at scale is a question the launch does not answer.
Muse Spark 1.1 and the Strategy Reversal
The upgrade is powered by Muse Spark 1.1, the model Meta released on July 9 from its Superintelligence Labs group. Meta describes Muse Spark as an agentic model built for planning and tool use, with a one-million-token context window and the ability to orchestrate sub-agents across applications, according to Unite.AI. The same model powers a new Meta Model API, now in public preview for US developers.
The strategic shift is significant. Meta had deliberately positioned its assistant around entertainment and social connection rather than productivity, a bet The Verge’s Alex Heath reported as a conscious decision to play to Meta’s strength in holding attention instead of chasing ChatGPT. This update abandons that line entirely.
The Consumer Agent Race
OpenAI, Google, and Anthropic have each pushed their assistants toward scheduled task execution and agentic workflows over the past six months. Meta’s entry positions Muse Spark 1.1 as a direct competitor on the same terrain. The competitive signal: Meta’s “personal superintelligence” pitch, the framing CEO Mark Zuckerberg has attached to the company’s broader AI spending, now needs productivity capabilities to be credible alongside ChatGPT Work, Gemini, and Claude Cowork.
The features are rolling out in select markets through the Meta AI app and meta.ai, with WhatsApp integration planned for the coming weeks.