Google Unveils Gemini Spark, an AI Assistant That Can Work Autonomously

Google has announced Gemini Spark at its annual developer conference, Google I/O 2026, held from 19 May 2026 US time. Gemini Spark is a personal AI agent that can assist users around the clock, 24/7. According to Google, Spark transforms Gemini from a mere AI assistant into an ‘active partner’ capable of carrying out tasks on behalf of users.

The capability is supported by the Gemini 3.5 AI model and is integrated with a range of applications within Google Workspace, such as Gmail, Docs, and Slides. Spark can be trained to perform a variety of tasks; for example, creating a list of important agenda items in Gmail, or writing a summary update from a long thread of emails that may still be under discussion.

Spark can also be programmed to complete several related tasks. For instance, instructing this AI agent to review meeting notes in chat and email, produce a neat report in Google Docs, and draft an email that can be sent together with that report.

The AI agent is optional within Gemini. Users can also set which apps can be integrated with Spark. In addition to Google Workspace, Spark can connect with third‑party apps such as Canva, Instagram, and OpenTable. Google promises more connected apps in the near future.

In the initial phase, Spark will be released to a number of testers, then to users of Google AI Ultra beta next week. In addition to the features above, Google will add more features to Spark in the coming weeks. These include the ability to send text messages and emails, as well as operating the user’s browser.

Not only new features, Google also promises Spark will be available in the Gemini desktop app in the US summer of 2026, roughly June–September. This means users will be able to command Spark to access documents and carry out tasks from a desktop or laptop computer, as reported by KompasTekno from Engadget.