Meta AI starts acting on your behalf
Powered by Muse Spark 1.1, Meta AI now runs recurring tasks, connects to email and calendar, and delivers daily briefings without re-prompting.
Meta began rolling out a productivity overhaul of Meta AI on July 24, built on Muse Spark 1.1, the agentic model Meta’s Superintelligence Labs group released July 9. The chatbot can now create and follow multi-step plans without repeated prompting, connect to Google Calendar and Gmail, generate presentation slides from research, run web-based research and synthesize the results, and deliver daily briefings that flag calendar conflicts. Users can set up a recurring task once, a weekly meal plan or an afternoon trend update, and the assistant runs it going forward without being asked again; it also accepts real-time course corrections mid-task and stores everything it produces in one place.
The rollout starts in select markets through the Meta AI app and meta.ai, with expansion to more countries and to WhatsApp planned for the coming weeks. Meta says users keep control over how the assistant behaves and that incognito chats stay available for private conversations. The company frames it as a step toward “personal superintelligence”:
“Meta AI is starting to take action on your behalf.”
It lands the same week as Amazon’s Alexa+ automation build-out: consumer assistants are moving from answering questions to holding calendar and inbox access and acting without a human confirming each step. The convenience is real, but so is the surface area: before connecting Meta AI to a primary email or calendar, check what it’s authorized to send or delete on your behalf, not just what it can read.