Google is preparing to roll out its most ambitious AI agent yet, dubbed “Gemini Spark.”
The always-on assistant aims to challenge Anthropic’s Claude Cowork by automating multi-step tasks across apps without manual supervision.
Found by messing around with the latest beta version (version 17.23) of Google’s app, this new feature allows users to have an “Agent” tab inside Gemini. Instead of behaving like regular chatbots that only answer questions, Spark is capable of doing tasks like cleaning up Gmail spam, compiling meeting summaries from various documents, and generating custom news summaries on its own.
Leaks suggest Spark lives inside Gemini’s overflow menu. When turned on, it can create “skills” to automate recurring tasks, much like Claude’s project feature.
Spark works with information collected from related apps, conversations, activities, geographic information, and even websites when the user is logged in.
Advanced capabilities could include controlling the Chrome browser like a self-driving car and being able to access documents on devices. But the early signs show that Spark can’t control an entire computer yet, unlike OpenClaw and Claude Cowork.
Despite offering significant results, Google cautions that the model remains “experimental.” The disclaimer is visible in the beta notes that although Spark requests permission before sensitive actions, it may have chances to share information or make purchases without the user’s permission.
