Google launches Gemini Spark AI agent for autonomous task automation

Google introduced Gemini Spark, an artificial‑intelligence agent that can operate independently across the company’s ecosystem. Unlike traditional assistants that respond only to direct commands, the agent interprets broad goals, breaks them into executable steps and runs them in the background. It relies on three components—Tasks for actions within Google services (Gmail, Docs, Drive), Skills for personalized behaviors, and Schedules for time‑based execution—built on the Gemini 3.5 Flash language model and the Antigravity infrastructure.

Gemini Spark also connects to external platforms such as Canva, OpenTable and Instacart, allowing users to request designs, restaurant reservations or other services without opening the apps. The launch marks Google’s push to turn conversational AI into a full‑featured productivity operator.

Separately, a review of 15 prompts tested Gemini’s ability to organise a 15‑day trip from São Paulo to Rio de Janeiro. The prompts generated a detailed itinerary, budget estimate and suggestions for attractions, hotels and transport. Reviewers noted the plan’s logical structure but flagged occasional timing and weather assumptions that require human verification. The test demonstrates Gemini’s practical applications beyond internal automation, extending to consumer‑focused planning tasks.