Google Launches Gemini Agent With Spend Caps

Google Cloud launched Gemini agent, enabling enterprises to delegate autonomous tasks across channels like Slack and Microsoft 365 while enforcing strict departmental budget caps.

Although per-token prices have dropped by approximately 98% since 2024, surging AI volumes have led to uncontrolled billing. The new system uses "smart routing" to assign simple tasks to Gemini Flash and complex long-horizon work to Argon or Anthropic Claude, balancing performance against cost.

The agent maintains persistent memory and can spawn sub-agents with dedicated email addresses for collaboration. Administrators can set hard spending limits in the Cloud Billing Console; if a cap is triggered, the agent pauses until manual approval overrides the budget. Current integrations are limited to select third-party apps, and real-world cost savings remain to be validated by users.