Google is expanding the role of its Gemini AI assistant towards more independent, agent-like operations.
At its I/O 2026 developer conference, the company introduced a series of updates aimed at transforming Gemini from a passive question-and-answer tool into a constantly working digital assistant in the background. New features include the Daily Brief agent, which provides daily summaries, and the 24/7 personal AI agent Gemini Spark.
The task of Daily Brief is to serve as a kind of morning starting point, compiling a personalized status overview for the user. The agent automatically reads connected applications in the background, such as Gmail and Google Calendar, and based on these, quickly forms a scannable summary of urgent messages, upcoming events, and, for example, tracking information for incoming packages. Daily Brief is not limited to mere summarization but also aims to prioritize content according to the user's goals and suggest next steps. The user can guide the agent by providing feedback on the summaries it produces. Daily Brief will initially be available only in the United States and is limited to Google AI Plus, Pro, and Ultra subscribers.
However, the most important of Gemini's updates is Gemini Spark, which Google describes as a 24/7 personal AI agent. According to Google, Spark signifies a significant change in Gemini's nature: it not only answers questions but acts as an active partner that performs concrete tasks on behalf of the user and within the rules defined by them. Spark is based on the new Gemini 3.5 model and can handle complex task chains even when all of the user's devices (phone, computer, etc.) are turned off.
Spark is deeply integrated into the Google Workspace environment, including Gmail, Docs, and Slides applications. For example, a user can ask it to filter emails related to children's school from their inbox, highlight important deadlines, and compile a daily summary of them, which is sent to both themselves and their spouse. Recurring tasks can be set for the agent, such as a monthly review of credit card statements for hidden subscriptions, or longer workflows can be built where Spark combines information gathered from various sources, produces ready-made documents, and drafts related emails. The goal is for users to be able to outsource time-consuming but repetitive aspects of knowledge work to AI.
Spark's operations are not limited to Google's own services, as the company is currently expanding so-called MCP connections to third-party applications. Thanks to these integrations, Spark will be able to not only retrieve information but also handle practical matters, such as making a restaurant reservation or creating shopping lists for groceries. In the coming weeks, more features have been promised for the agent, including the ability to send text messages and emails directly through Spark, and the option to use the user's browser as part of task execution.
Since it is largely an independently operating agent, Google emphasizes the importance of control and security. Spark is opt-in by default: the user decides whether to enable it and to which applications it can connect. According to the company, the agent is designed to explicitly ask for permission before high-risk actions, such as making payments or sending messages. This aims to minimize situations where AI would make significant decisions for the user without clear approval.
Gemini Spark will initially be released to a limited group of trusted testers, after which it will be rolled out as a beta version to Google AI Ultra subscribers in the United States. Broader availability and a more detailed country-specific distribution of features are currently open. For example, strict privacy laws in Europe may limit Spark's functionality in some respects.
The task of Daily Brief is to serve as a kind of morning starting point, compiling a personalized status overview for the user. The agent automatically reads connected applications in the background, such as Gmail and Google Calendar, and based on these, quickly forms a scannable summary of urgent messages, upcoming events, and, for example, tracking information for incoming packages. Daily Brief is not limited to mere summarization but also aims to prioritize content according to the user's goals and suggest next steps. The user can guide the agent by providing feedback on the summaries it produces. Daily Brief will initially be available only in the United States and is limited to Google AI Plus, Pro, and Ultra subscribers.
However, the most important of Gemini's updates is Gemini Spark, which Google describes as a 24/7 personal AI agent. According to Google, Spark signifies a significant change in Gemini's nature: it not only answers questions but acts as an active partner that performs concrete tasks on behalf of the user and within the rules defined by them. Spark is based on the new Gemini 3.5 model and can handle complex task chains even when all of the user's devices (phone, computer, etc.) are turned off.
Spark is deeply integrated into the Google Workspace environment, including Gmail, Docs, and Slides applications. For example, a user can ask it to filter emails related to children's school from their inbox, highlight important deadlines, and compile a daily summary of them, which is sent to both themselves and their spouse. Recurring tasks can be set for the agent, such as a monthly review of credit card statements for hidden subscriptions, or longer workflows can be built where Spark combines information gathered from various sources, produces ready-made documents, and drafts related emails. The goal is for users to be able to outsource time-consuming but repetitive aspects of knowledge work to AI.
Spark's operations are not limited to Google's own services, as the company is currently expanding so-called MCP connections to third-party applications. Thanks to these integrations, Spark will be able to not only retrieve information but also handle practical matters, such as making a restaurant reservation or creating shopping lists for groceries. In the coming weeks, more features have been promised for the agent, including the ability to send text messages and emails directly through Spark, and the option to use the user's browser as part of task execution.
Since it is largely an independently operating agent, Google emphasizes the importance of control and security. Spark is opt-in by default: the user decides whether to enable it and to which applications it can connect. According to the company, the agent is designed to explicitly ask for permission before high-risk actions, such as making payments or sending messages. This aims to minimize situations where AI would make significant decisions for the user without clear approval.
Gemini Spark will initially be released to a limited group of trusted testers, after which it will be rolled out as a beta version to Google AI Ultra subscribers in the United States. Broader availability and a more detailed country-specific distribution of features are currently open. For example, strict privacy laws in Europe may limit Spark's functionality in some respects.
