Google says Gemini Enterprise now autonomously orchestrates 200+ models

Google on April 22 introduced Gemini Enterprise Agent Platform, a new Google Cloud product that expands Vertex AI into a broader system for building and managing AI agents. (cloud.google.com) Google described it as the “evolution of Vertex AI,” combining model selection, tuning, and agent-building with new tools for orchestration, DevOps, integration, and security. (cloud.google.com) The platform’s documentation says it supports the full AI lifecycle, from access to more than 200 foundation models to deployment and management of enterprise agents. Google’s launch materials highlight Gemini 3.1 Pro, Gemini 3.1 Flash Image, and Lyria 3 among the models available. (docs.cloud.google.com) (blog.google) Google also said Gemini Enterprise customers now want agents that can execute multi-step workflows for “hours or days,” not just answer prompts or complete one-off tasks. The company tied that shift to new product features including long-running agents and agentic collaboration spaces. (cloud.google.com) Memory is part of that pitch. Google’s new Memory Bank documentation says agents can generate long-term memories from user conversations that persist across sessions, while Gemini Enterprise documentation says admins and users can control personalization and memory settings. (docs.cloud.google.com 1) (docs.cloud.google.com 2) Google is also pitching openness. Its Gemini Enterprise app page says organizations can use Google-made, third-party, and internally built agents in one place, and cites support for open standards such as the Agent2Agent protocol for interoperability. (cloud.google.com) That matters because the launch recasts Gemini Enterprise as more than a workplace chatbot. Google’s April 22 product post says the service is now an “end-to-end system for the agentic era,” aimed at delegating business outcomes to software agents rather than just assisting employees with prompts. (cloud.google.com) Google’s own framing is aimed at large companies with compliance demands. The company said enterprises need agents with identity, registry, and gateway controls so they can be traced, monitored, and managed inside secure environments. (cloud.google.com) Release notes published this week show Google is already adding pieces around that governance layer, including agent identity visibility for administrators and a pre-built Deep Research Agent on the new platform. (docs.cloud.google.com 1) (docs.cloud.google.com 2) The immediate change is that Google is no longer selling Gemini Enterprise only as an assistant that answers questions. It is now selling a managed stack for companies that want AI systems to run tools, coordinate models, remember context, and keep working after the first prompt. (cloud.google.com) (docs.cloud.google.com)