Google launches Gemini enterprise agent

Enterprise AI is moving past the “pick a model and ship a chatbot” phase. The harder problem now is managing fleets of agents that call tools, touch data, and act across real business systems without turning into a governance mess. That is the gap Google is trying to close with Gemini Enterprise Agent Platform, unveiled at Google Cloud Next ’26 on April 22. Basically, Google took Vertex AI, folded in new agent tooling, and repositioned the whole stack around building, deploying, governing, and optimizing agents at enterprise scale. (cloud.google.com) ### What actually launched? Google launched Gemini Enterprise Agent Platform as the successor to Vertex AI, not as a side product. The company describes it as a single destination for technical teams to build agents, connect them to enterprise data and tools, deploy them, and keep them under policy and security controls. The rename matters because it signals a shift from model-centric infrastructure to agent-centric infrastructure. (cloud.google.com) ### Why is Google changing the name at all? Because “Vertex AI” sounded like a toolbox for models, while Google now wants to sell a full operating layer for autonomous workflows. The new platform bundles the model-building pieces customers already used with added orchestration, DevOps, integration, and security fea(cloud.google.com)e around how agents are built and supervised. (cloud.google.com) ### What’s inside the platform? The core pieces are build, govern, and run. Google’s docs and launch posts point to agent-building tools, an Agent Development Kit, orchestration, and a governance layer that includes visibility into agents, endpoints, and Model Context Protocol servers across an organization. There (cloud.google.com)at exists instead of losing track of dozens of semi-connected automations. (docs.cloud.google.com) ### Why does MCP show up here? Because once agents start using external tools, you need a standard way to expose those tools and track them. Google is explicitly building governance around MCP servers, which is notable — it means the company is not pretending enterprise AI will stay inside one vendor’s walls. The platform is leaning into open-(docs.cloud.google.com)f the sales pitch. (docs.cloud.google.com) ### Is this only about Google models? No. Google says the platform offers access to more than 200 models, including Gemini-family models and third-party options. That matters because large companies rarely want to bet everything on one model vendor. The platform play is stronger if Google can be the place where mixed-model agent systems get built and controlled, even when some of the brains come from elsewhere. (letsdatascience.com) ### What’s the enterprise angle? Governance is the real headline. Google keeps emphasizing identity, security, policy, auditability, and centralized oversight. That sounds boring next to model benchmarks, but turns out it is exactly what big companies need once agents can trigger actions instead of just generate text. An agen(letsdatascience.com)t, and prove what it did. (docs.cloud.google.com) ### Why does this matter now? Because the market is maturing. Last year’s excitement was about raw model capability. This year’s fight is about who owns the layer above the model — the one that handles orchestration, connectors, identity, observability, and compliance. Google is trying to make that layer look like cloud infrastructure, not a loose collection of AI demos. (cloud.google.com) ### Bottom line? Google’s launch is really a bet on where enterprise AI spending goes next. If companies buy managed agent platforms instead of just API access, the winner will not be the lab with the flashiest model alone. It will be the vendor that makes autonomous software feel governable enough to run inside a real business. (c([cloud.google.com)gent-platform))