Google Cloud introduces the Gemini agent.

Google Cloud has unveiled Gemini Agent, an autonomous AI assistant capable of generating code and completing work across various devices. This agent can be accessed through multiple platforms, including web, mobile, desktop, Google Workspace, Microsoft 365, Slack, and even third-party applications. Google Cloud Chief Executive Thomas Kurian highlighted Gemini's ability to handle knowledge work, create media, and transform ideas into finished products.

The agent operates by receiving objectives rather than specific instructions, learning from user interactions, and maintaining context across all communication channels. It also has the flexibility to choose from different models based on the task, ranging from Gemini Flash for simple tasks to powerful models like Argon for complex work.

Gemini observes and incorporates a company's unique knowledge, grounding answers in its institutional data. Users can build custom skills as reusable templates, and the system learns from every interaction to adapt its understanding of users' workflows. Google has integrated specialized skills for data science and engineering, enabling Gemini to generate code, create notebooks, and train models.

Despite a significant drop in per-token prices since 2024, enterprise AI usage has surged. Google Cloud is offering flexible spending options to manage costs. Gemini's multimodel orchestration ensures optimal model selection for each task, and smart routing automatically assigns work to the most cost-effective model. Users can set spending limits, and the agent will pause when those limits are reached, requiring user approval to continue.

This granular cost tracking allows businesses to allocate AI expenses to specific departments, facilitating budget planning and audits.