Google's New Gemini Agent for Business

On 8 October 2026, Google Cloud announced a new Gemini agent designed to move workplace AI beyond answering questions and towards completing business tasks. The company describes it as a single, universal AI agent capable of handling knowledge work, producing content, writing code and working across business applications. For UK small and medium-sized businesses, the announcement raises an important question: could AI become a practical digital colleague rather than simply another tool employees have to operate?

According to TechCrunch, Google is initially focusing on businesses, with a consumer rollout planned for later. The announcement includes substantial promises about automation, security and cost management, but important commercial details remain unanswered. Here is what Google has announced, how its agent could work inside familiar business tools, and what smaller UK organisations should consider before deciding whether to adopt it.

What are the key facts about Google's Gemini agent?

What is the Gemini agent and how does it work?

The central change is the difference between asking AI to perform individual actions and giving it responsibility for an outcome. Thomas Kurian, CEO of Google Cloud, explained that users give Gemini objectives rather than instructions. Instead of manually directing every stage, a person can describe the desired result and allow the agent to plan the work, select tools, use relevant business information and complete the task.

Google Cloud says the agent combines conversational assistance, autonomous task execution and code generation within one interface. It can also schedule work or respond to events. For a business owner, the significant distinction is that Gemini is designed to operate across several activities rather than simply generate text in response to a prompt.

Another feature is persistent execution. According to Google Cloud, Gemini runs in the cloud and maintains context across devices and applications. Tasks lasting hours or days can continue after a user closes their laptop. The system can also create temporary specialist sub-agents to tackle different parts of a complex assignment, coordinating their work as necessary.

Google describes three underlying capabilities: tools that connect Gemini to business systems, skills that teach it repeatable workflows, and memory that preserves relevant context. Teams can develop reusable skills for their own processes. The intended result is an agent that becomes more familiar with an organisation's ways of working, rather than starting every assignment without context.

Does Gemini work inside Gmail and Google Workspace?

Yes. Google Cloud says the Gemini agent works directly within Gmail, Drive, Docs, Slides, Sheets, Chat and Calendar, carrying consistent memory, skills and controls between them. This matters because many business activities involve several applications, even when the final task appears straightforward.

Google describes three ways the agent operates inside Workspace. The first is personal assistance. Gemini can use relevant calendars, conversations, team relationships and project information to coordinate work. In Google's example, someone can request a meeting with their usual regional event leads, and the agent identifies participants and begins arranging a suitable time.

The second is proactive delegation. Google says Workspace Intelligence can recognise potentially delegatable work, such as an email requesting a project update in presentation form. The employee can then delegate the assignment to Gemini rather than performing every stage personally. The system can also prioritise important inbox messages and explain their relevance.

The third is team collaboration through a coworker agent. Google says businesses can describe a role and create a dedicated agent with its own Workspace identity, email address, calendar, Drive storage and company directory presence. Colleagues can email it, mention it in Chat or assign work through document comments.

For example, Google describes an events coordinator agent drafting a launch readiness document when requested by a marketing manager. The resulting document can be shared back with the team. Importantly, Google says coworker agents operate under their own identities and access only the context shared with them through applicable permissions.

Which business tools and AI models can Gemini use?

Google is positioning the agent as a system that connects to existing business software, not just its own applications. Google Cloud lists Microsoft Office, Teams, Slack, Confluence, Salesforce, ServiceNow, Git, Jira, BigQuery, Databricks, Postgres and Snowflake among supported connections. The company also describes integration with Model Context Protocol servers inside or outside an organisation's network.

Access is intended to extend beyond Workspace. According to TechCrunch, Gemini can be used through mobile devices, desktop computers, command-line environments and business applications. Google Cloud also describes integration into third-party software without requiring a dedicated user interface.

Model choice is another significant feature. Rather than requiring every assignment to use the same underlying AI model, Google says the agent can select one suited to the task. Its current approach includes Google's Gemini models and Anthropic's Claude models, while private and open models are planned for the future.

Google argues that selecting models according to workload can improve quality and reduce unnecessary spending. For smaller organisations considering automation, the broader practical question is whether connecting existing processes and tools produces useful results without adding excessive complexity.

How secure is Gemini and can businesses control costs?

Giving an AI agent permission to act inside company systems makes governance essential. Google Cloud says every agent receives a distinct, cryptographically attested identity with least-privilege permissions. Administrators can apply role-based access controls, while actions are recorded in audit trails attributed to the agent rather than an employee.

Google also describes an Agent Sandbox with its own network boundary. Traffic between agents and external systems passes through Agent Gateway, which enforces organisational policies. The company says administrators can establish restrictions on information agents must not access, with those rules applied across their organisation.

These are Google's descriptions of its security architecture, not independent guarantees that every possible deployment will be risk-free. Businesses still need to examine the permissions, information and workflows involved in their own proposed use.

Cost management is another part of the announcement. Google Cloud describes multi-model orchestration, Smart Routing and real-time spending caps. It says businesses can establish hard project spending limits in the Cloud Billing Console. When a cap is reached, the relevant agent pauses until someone chooses to resume it.

However, spending controls are not the same as published subscription prices. Neither Google Cloud nor TechCrunch gives a specific small-business price. A UK company therefore cannot calculate its expected subscription commitment from this announcement alone.

How widely is Google AI used in business already?

Google and TechCrunch present substantial adoption figures, although these should not be confused with confirmed adoption of the newly announced agent. TechCrunch reports that Google CEO Sundar Pichai said Gemini has more than one billion monthly active users and that nearly 90% of Fortune 100 companies use Gemini Enterprise at work.

Google Cloud also claims that nearly 80% of its customers use its AI products, while almost 500 Google Cloud customers each processed more than one trillion tokens over the preceding year. Those figures illustrate the scale Google reports across its wider AI business, rather than the number of small firms already using this particular agent.

Google Cloud highlights several customer examples. It claims Bradesco reduced document review time from one hour to five minutes through Gemini Enterprise workflows, while Bunnings saved staff half a million hours of administrative work using an internal agent. These are Google's own customer-result claims, not independently established forecasts for other businesses.

There is also a UK example. Google Cloud says Arden University, a provider of flexible online and blended learning, is rolling out Gemini Enterprise to more than 30,000 students and staff. The initiative is intended to support personalised learning and workplace skills. It demonstrates a UK organisational deployment of Gemini Enterprise, but does not establish availability or performance for a typical small business.

What is not yet known for UK small businesses?

Despite the breadth of the announcement, several important purchasing questions remain unanswered. Neither Google Cloud nor TechCrunch provides UK availability dates for the new Gemini agent. They do not specify small-business pricing, subscription plan names or a clear purchasing route for smaller UK organisations.

TechCrunch explicitly says Google is concentrating on businesses first and intends to reach consumers later. That does not establish exactly which businesses will receive access first, or when a UK small company will be able to adopt the new capabilities.

The published examples also concentrate heavily on large organisations and enterprise-scale workflows. They demonstrate what Google says its technology can achieve in particular circumstances, rather than a guaranteed return for smaller companies with different budgets, data and processes.

Until clearer commercial information emerges, UK owners should distinguish between announced technical capabilities, examples from Google's existing enterprise customers and features they can actually purchase and deploy themselves.

What should a UK small business do now?

There is no reason to treat Google's announcement as an instruction to immediately replace existing processes. A more useful approach is to assess where delegated automation could genuinely help. The following readiness checklist represents general business good practice, not requirements imposed by Google.

1. Tidy your business data. Identify outdated documents, duplicate records and inconsistent information before allowing automation to rely on them. Clear, organised information makes useful workflows easier to design and review.
2. Review existing permissions. Check who can access shared folders, customer records and business applications. Consider whether those access rights remain appropriate before introducing additional automated users.
3. Choose one pilot task. Identify a repeatable activity with a defined result, such as preparing an internal project summary. Select something manageable before considering more complicated processes.
4. Appoint a named owner. Give one responsible employee oversight of the pilot, including permissions, approval decisions, exceptions and the quality of completed work.
5. Establish a spending limit. Decide what the business is prepared to invest in testing and ongoing operation. Review any available platform spending controls before committing to automation.
6. Define what data an agent may touch. Document the approved information sources, restricted materials and systems involved. Be clear about actions requiring human approval.
7. Brief your staff. Explain the intended workflow, who remains accountable and how employees should review outputs or raise concerns about unexpected behaviour.
8. Measure the results. Compare the pilot with the previous process, including time spent, quality, corrections and operating costs. Expand only where the evidence supports doing so.

Is the new Gemini agent available to UK businesses?

Google Cloud announced the agent on 8 October 2026, and TechCrunch says businesses are the initial focus. However, neither publisher provides a specific UK availability date or confirms when smaller UK organisations can obtain it. Businesses should therefore avoid treating the announcement as confirmation of immediate access.

Can Gemini send emails and edit Google Docs?

Google Cloud says Gemini works directly within Gmail and Docs alongside other Workspace applications. Its examples include coordinating meetings through email, drafting documents and suggesting edits in document comments. Coworker agents can have their own email addresses and identities, allowing their actions to be distinguished from those of employees.

How much will Google's new Gemini agent cost?

Neither Google Cloud nor TechCrunch states subscription prices, small-business charges or plan names for the newly announced agent. Google describes Smart Routing, model selection and real-time project spending caps as cost-management features, but these do not establish what a UK small business would pay to use the service.

Will Gemini replace employees in small businesses?

Google Cloud presents Gemini as both a personal assistant and a team member capable of completing delegated objectives. Its examples include coordinating meetings, preparing documents and managing multi-step work. The announcement does not establish that employees will be replaced. For small firms, a practical starting point is to test specific tasks while keeping people responsible for decisions and oversight.