The news
Google Cloud introduced Gemini Agent on October 8, 2026, an AI assistant designed to carry out work for employees across many of the applications companies already use, SiliconANGLE reported. Google Cloud CEO Thomas Kurian described the idea as giving the system objectives, not instructions, and returning to finished work.
Gemini Agent can run on the web, mobile and desktop, and can be reached from a command line, Google Workspace, Microsoft 365, Slack and third-party apps, according to SiliconANGLE. It keeps a shared memory and context across those channels and, the outlet reported, learns from interactions over days.
The integrations list is broad. SiliconANGLE reported connections to Confluence, Microsoft Office and Teams, Git and Jira, Salesforce and ServiceNow, and databases including BigQuery, Databricks, Postgres and Snowflake, as well as desktop files.
Two features stand out for IT teams. The agent can create sub-agents, each with its own identity and an email address on an agents subdomain of the customer's domain. And it can choose a model per task: Google's Gemini Flash for simple jobs and its Argon frontier model for complex work. Anthropic's Claude models are available at launch, with other private and open models to follow, SiliconANGLE reported. Google is also offering custom skills, reusable workflows and domain-specific skills, starting with data science.
On cost, Google pointed to smart routing between models, hard spend limits enforced in its Cloud Billing Console and per-project cost tracking so departments can be charged back. SiliconANGLE reported that flexible spending options announced in August took effect on October 8. Pricing for Gemini Agent itself was not detailed in the report.
The numbers
- Launch date
- October 8, 2026
- Per-token price drop Google cites since 2024
- about 98%
Why CEOs should care
For CIOs, Gemini Agent is a bid to become the single front door for workplace AI, reaching into Microsoft 365 and Slack as well as Google's own tools. If you already pay for several overlapping assistants, ask whether one agent across systems could replace some of them, and what you would lose in depth for each app.
CISOs should focus on the sub-agent identities. Agents with their own email addresses and access rights are effectively new digital workers. Before rollout, decide who approves an agent's permissions, how its actions are logged and how it is shut off. Treat each agent like a service account with an owner.
CFOs get controls they have been asking for: hard spend caps and per-project tracking. Set limits before pilots begin, because agents that work on goals over days can run up model usage quickly. Ask Google for clear pricing on the agent and on the third-party models, such as Claude, that it routes to.
The bigger picture
The launch builds on Google's agent push from earlier in the year. At Google Cloud Next in April, Kurian positioned Gemini Enterprise as a control layer for corporate AI agents and said the company's advantage lies in combining models, infrastructure, data, security and workplace tools, IT Brief reported. Customers cited then included KPMG and Signal Iduna.
Microsoft, OpenAI, Anthropic and Salesforce are all chasing the same budget for agents that do work rather than answer questions. Google's choice to include rival models such as Claude suggests it sees control of the workflow, more than the model, as the prize.
What’s next
Watch for Gemini Agent pricing, the timing of further third-party models and early customer results. Companies should run limited pilots with spend caps and agent identity rules in place before broad rollout.
What “Fact-checked” means
Fact-checking means testing a story’s facts against the evidence before it is published. This story went through at least two separate checks before this version was published.
- What we checked
- Its names, figures, dates, job titles, quotes and who said what were checked against the story’s sources, including its main source where it could be opened. The headline was checked for accuracy and overstatement.
- How
- A first check reviewed the whole story. If it passed, a second, skeptical check went back to the sources to look for mistakes in the most important facts. If a check flagged the story, it was edited to fix the problems found, and a separate re-check then reviewed the whole story again.
- Who
- The checks are made by our newsroom, as steps kept separate from the writing, under rules set by our editor, Hussein Mukhtar. A story the checks still flag is held for the editor, who decides whether it is fixed, published or dropped.
- If something is wrong
- “Fact-checked” does not mean error-free. If a material error is found after publication, we correct the story and add a note saying what changed. Report an error
Companies in this story
