Google Launches Gemini Agent for Businesses at Gemini at Work
Google Cloud announced the Gemini agent on 8 October 2026 at Gemini at Work, a business-first work agent that plans tasks from a single prompt box before a later consumer rollout.
PromptCrates Editorial
Staff Writer

Google Cloud announced the Gemini agent on Thursday, 8 October 2026, at its Gemini at Work event, describing it as a universal agent for work that starts from a single prompt box. Google's blog post says the agent plans work, uses skills and tools, connects to customers' business systems, and returns finished results inside documents, inboxes and developer environments. TechCrunch reported that Google will roll the agent out to businesses first and to consumers later, after CEO Sundar Pichai said Gemini has more than 1 billion monthly active users and that nearly 90% of Fortune 100 businesses already use Gemini Enterprise.
What the Gemini agent can do at work
Thomas Kurian, CEO of Google Cloud, told the event that the agent can be given objectives rather than only step-by-step instructions, according to TechCrunch. That framing matches Google's own summary: the agent chooses a model for the job, applies built-in cost controls, and is meant to carry the security, administration and governance enterprises require. Users can attach files, folders or project bundles that combine files and skills, and the agent can connect to systems such as Google Workspace, Microsoft 365, Slack, Jira, Confluence, Git, BigQuery, Databricks, Postgres and Snowflake. TechCrunch also reported that it can work with any Model Context Protocol server inside or outside the company network.
By default the agent picks a model, but people can override that choice. Google said the model picker will include third-party models starting with Anthropic's Claude, with open-source and other private models planned later. The agent gets its own Workspace account, complete with an email address and organisational context about teams, time zones, approvers and calendars. People can tag it, email it, share with it or add it to a group chat, and its actions write an audit trail attributed to the agent rather than to a human coworker. A tasks inbox shows thinking, subagent delegation, skill loading, code and progress.
Why Google is starting with businesses
Pichai argued that starting with enterprises lets Google solve harder problems around security, scale and performance before a consumer launch, TechCrunch reported. Early testers named in that coverage include sportswear brand On, Shopify and PayPal. Gemini Enterprise customers cited include BNP Paribas, Bradesco, Merck, Orange Spain, Santee Cooper, SOMPO, Ulta Beauty and Wesfarmers. Google also pointed to flexible spending options such as multi-model orchestration, smart routing and real-time spend caps.
The enterprise-first path sits on top of earlier Gemini Enterprise packaging that PromptCrates has tracked, including Antigravity Gemini Enterprise subscriptions and vertical pushes such as legal and financial services. Those earlier stories focused on seat packaging and industry workflows; Thursday's announcement is about a single agent surface that is supposed to own multi-step work rather than answer one chat turn at a time.
Access is broad once a business has it. TechCrunch listed iOS and Android, Windows and Mac desktops, the command line, Google Workspace, Microsoft 365, ServiceNow and Slack as places where the agent can appear. That distribution matters for companies that already live in mixed Google and Microsoft stacks, because the connectors Google listed span both.
How this fits the wider agent race
Agent products have been shifting from chat to task ownership across the industry, a shift TechCrunch linked to consumer agents such as Meta's Muse, messaging agents and ChatGPT's Dots. Google's pitch is that Gemini already has consumer scale and Fortune 100 penetration, so the company can harden governance in workplaces before opening the same pattern to individuals. The agent-as-coworker model, with its own mailbox and audit trail, is a concrete answer to the identity problem other vendors have approached through sandboxes and policy layers, including Microsoft's Agent Governance Toolkit in the broader agent-governance conversation PromptCrates has covered.
Cost control is part of the same pitch. Google's blog stresses built-in cost controls; TechCrunch details spend caps and smart routing across models, including the option to call Claude. For buyers, that is as important as feature lists, because long-running agents that read large corpora can become expensive quickly if every step bills a frontier model.
What to watch after Gemini at Work
Several practical questions remain after the keynote. Google has not, in the materials reviewed here, published a consumer launch date, a detailed pricing card for the new agent, or independent evaluations of how often the agent finishes multi-step jobs without human rescue. The model picker that includes Claude is a notable openness move for Google, but the timeline for open-source and private models is still described only as future expansion. Buyers will also want clarity on how audit trails, MCP connections and third-party models interact with existing data-loss-prevention rules.
For now, the verifiable news is narrower and dated to 8 October 2026: Google Cloud introduced a Gemini agent aimed first at businesses, presented it as a single prompt-driven worker with enterprise connectors and governance, and said consumer availability comes later. Organisations already on Gemini Enterprise can treat Thursday as the start of a product cycle to test objectives-based prompts, inbox supervision and multi-model routing on real internal workflows before any wider consumer story arrives. Teams that already orchestrate agents across Workspace and Microsoft 365 should also pressure-test the audit trail claim: when an agent emails colleagues or edits shared files, IT needs to see those actions as agent activity rather than as the human who summoned it. Google's materials emphasise that attribution model, and TechCrunch's account of the Workspace-native identity is the clearest public description so far of how that separation is supposed to work in day-to-day collaboration.
- Google: Google Cloud introduces the Gemini agent
- TechCrunch: Google brings agentic AI to Gemini, starting with businesses
- PromptCrates: Google Antigravity Gemini Enterprise subscriptions
- PromptCrates: Google Gemini Enterprise for legal and financial services
- PromptCrates: Microsoft Agent Governance Toolkit on GitHub


