How you can use the Gemini agent
Instead of juggling separate tools, individuals can give the Gemini agent an objective and return to finished work. It can call on temporary sub-agents for multi-step jobs, and companies can create “coworker agents” with set roles, their own email addresses and storage, and access only to information they are given. For each task, it picks a model, currently from Google’s Gemini family or Anthropic’s Claude, with others planned.
Giving an AI agent its own email address and access to company systems raises an obvious question: who keeps it in check? Google says each agent has its own identity, so its actions are logged under its name rather than an employee’s. Its traffic also passes through a gateway that applies company rules, such as keeping agents out of restricted documents (a concern with previous AI agents).
Cost is another worry. Google states that token prices have fallen 98% since 2024, yet heavier use still strains budgets. To keep spending on autonomous AI under control, administrators can set a budget cap for each project. Its agent pauses when that limit is reached.
Rival AI agents
The AI race is playing out in different ways across the industry. Meta’s Muse handles shopping and travel for consumers remains limited to the United States. OpenAI’s dots are rolling out on paid ChatGPT plans, though Pro subscribers in the UK and European Economic Area miss out for now. Microsoft added a persistent Autopilot agent to a redesigned Copilot on 25 September. Google’s route differs: one agent that works across Microsoft, Salesforce and ServiceNow systems and across competing models, rather than inside a single vendor’s suite.
Google’s strongest evidence so far comes from adoption of its existing products. It reports that nearly 80% of its cloud customers use its AI products and nearly 90% of Fortune 100 companies use Gemini Enterprise. The new agent, however, is still in private preview, leaving limited evidence of its own performance.