Google Cloud announced its Gemini agent for work at Gemini at Work 2026 on October 8, 2026. Google describes it as an agent designed to plan tasks, use tools and connect to business systems, then return completed work in documents, inboxes and developer environments.
What Google says the agent is designed to do
Google’s announcement presents the product as a unified work agent: a user starts with a prompt, and the agent plans the work, uses skills and tools, connects to business systems and delivers a finished result. The description also includes work in documents, inboxes and developer environments.
Launch coverage additionally reported that the agent can answer questions, handle tasks, create content and write code. Those functions fit the broader work-agent pitch, but they remain reported capabilities rather than independently tested results.
Reported integrations and model options
The agent is reported to work with Gmail, Docs, Sheets and Calendar, as well as Microsoft 365 and Slack. That mix would place it across Google’s workplace apps and tools from other providers.
Launch coverage also reported that the agent could draw on models from Google’s Gemini family and Anthropic’s Claude family. The announcement’s broader idea is to coordinate work across tools; the reported integrations and model choices add detail to that plan.
Industry versions described at announcement
On October 8, financial-services and legal versions were reported to be in preview. Versions for government, healthcare and retail were described as coming soon. These are sector-specific stages reported alongside the announcement, rather than a single availability status for every version.
Gemini work agent, Gemini Spark and Workflow Builder
The new work agent is distinct from Gemini Spark, Google’s separately described personal productivity agent. The Spark page concerns that personal product, not the Google Cloud work-agent announcement.
Gemini Enterprise Workflow Builder is different again: Google documents it as a way to create chat agents, either from a prompt or with a flow builder, then connect data sources and tools, preview the agent and publish it. That is an agent-building feature, not the unified work agent Google Cloud announced.
