Google rebrands Vertex AI as Gemini Enterprise Agent Platform with governance tools for managing agent fleets

Google has rebranded its Vertex AI developer platform as the Gemini Enterprise Agent Platform at Google Cloud Next, positioning it as an orchestration layer for enterprises deploying hundreds or thousands of AI agents.

Platform architecture

The platform is organized around four pillars: build, scale, govern, and optimize. Key components include:

• Agent Studio: Low-code interface for creating agents using natural language
• Agent Development Kit: Graph-based framework for orchestrating multiple agents working together
• Agent Registry: Central catalog of internal agents and tools
• Agent Marketplace: Pre-built agents from partners including Atlassian, Oracle, ServiceNow, and Workday

Performance specifications

According to Google, Agent Runtime delivers sub-second cold starts and provisions new agents in seconds. The platform supports long-running agents that can operate for hours or days on complex workflows like financial reconciliation or sales prospecting. A Memory Bank feature provides persistent, long-term memory across sessions.

Governance and security features

Google introduced several security tools:

• Agent Identity: Assigns each agent a unique cryptographic ID with defined authorization policies and auditable action trails
• Agent Gateway: Enforces security policies and protects against prompt injection, tool poisoning, and data leakage
• Agent Anomaly Detection: Flags suspicious behavior by analyzing agent intent before execution

Optimization tools

The platform includes Agent Simulation for stress-testing against synthetic interactions before deployment, Agent Evaluation for scoring live performance, and Agent Observability dashboards for tracing execution paths and real-time debugging.

Google Cloud CEO Thomas Kurian stated the shift from building individual agents to managing agent fleets represents "the next evolution of where we see this AI technology going."

Consumer-facing tools

The Gemini Enterprise app includes Agent Designer, allowing non-technical employees to create schedule- or trigger-based agents. An Inbox feature provides a central hub for monitoring agent activity with notifications categorized as "Needs your input," "Errors," and "Completed."

Google CEO Sundar Pichai cited internal adoption statistics showing 75 percent of all code at Google uses agentic workflows, though the full context of this statement was not provided in the announcement.

Infrastructure announcements

Google announced its eighth generation of TPU chips alongside the platform rebrand and referenced security updates through its Wiz acquisition, though specific technical specifications for the TPU chips were not disclosed.

What this means

Google is betting that enterprise AI adoption will create a management complexity problem as companies deploy agent fleets rather than individual models. The governance features—particularly cryptographic identity and anomaly detection—address real concerns about autonomous agents operating without oversight. However, the platform's success depends on whether enterprises actually face the "agent sprawl" problem Google is positioning to solve, or whether centralized platforms from competitors like Salesforce and ServiceNow capture this emerging market first.