Gemini Enterprise Agents

Google used Cloud Next on April 22 to turn Vertex AI into the Gemini Enterprise Agent Platform, a new system for building and policing fleets of workplace AI agents. (cloud.google.com) Google said the platform combines model selection, tuning, agent building, DevOps, orchestration and security in one product. It also said future Vertex AI services and roadmap updates will ship through Agent Platform instead of as a standalone service. (cloud.google.com) An AI agent is software that can take a goal like “review these contracts” or “prepare a sales brief,” pull in company data, use tools and keep working across multiple steps. Google’s pitch is that companies now need one control layer for those agents before they spread across departments and apps. (docs.cloud.google.com; cloud.google.com) Google built that control layer around retrieval, permissions and logs. Its documentation says the platform includes a Retrieval-Augmented Generation engine to ground answers in private data, Identity and Access Management controls for agent permissions, policy tools for monitoring inputs and outputs, and execution traces for debugging and audits. (docs.cloud.google.com) The model menu is broader than Google-only software. Google said Agent Platform offers access to more than 200 models through Model Garden, including Gemini 3.1 Pro, Gemini 3.1 Flash Image, Gemma 4 and Anthropic’s Claude models. (cloud.google.com) Google also tied the developer platform to employee-facing products. The Gemini Enterprise app is now the front door where workers can discover, run and share Google-made, custom and partner-built agents, with partners including Oracle, Salesforce and ServiceNow. (cloud.google.com; cloud.google.com) That same cross-app push is showing up inside Google Workspace and Chat. Google said “Ask Gemini in Chat” is a command line inside Google Chat that can produce finished work in the chat window, and Workspace Studio lets employees build and share no-code agents that connect to business apps. (workspace.google.com; workspace.google.com) Chrome is getting agent features too. Google said Gemini in Chrome can work through multi-step workflows with “auto browse,” while a new “Skills” feature lets users save repeatable prompts and run them on the page they are viewing. (blog.google; blog.google) ZDNET described the release as a tool for “managing fleets of agents at scale,” and highlighted Google’s emphasis on agent identity, simulation and secure deployment. That framing matches Google’s own message that the hard part is no longer just generating text, but supervising autonomous software across enterprise systems. (zdnet.com; cloud.google.com) Google’s closing argument is that enterprise AI will look less like a chatbot tab and more like a governed layer that sits across documents, chats, browsers and back-office software. The company is now selling the plumbing for that layer as much as the models themselves. (blog.google; cloud.google.com)