admin The platform marks a wider push by Google to reframe its enterprise AI business around agents that can do more than answer prompts. Under the new structure, Google is positioning Gemini Enterprise as the umbrella for building, deploying, governing and optimising agents across an organisation, while the underlying Agent Platform is presented as the evolution of Vertex AI for agent-based workflows. The offering is aimed at companies trying to move from AI pilots to systems that can carry out multi-step operational tasks across departments and software environments.
At the centre of the announcement is Agent Identity, which assigns every agent a strongly attested cryptographic identity based on the SPIFFE standard. Google says those identities are tied to the lifecycle of the agent and mapped directly to where it runs, allowing the agent to authenticate itself securely to cloud resources, servers, endpoints and other agents. The system is designed to reduce the reliance on shared credentials and long-lived keys, two longstanding weak points in enterprise security, while binding access tokens to the agent’s own certificates to make theft or misuse harder.
That matters because the commercial AI market is shifting away from single chat interfaces towards fleets of specialised agents handling procurement, coding, customer service, internal search, document workflows and finance operations. Once those agents begin accessing company systems, the question is no longer just how clever the model is, but who authorised the action, what data was touched, whether the tool stayed within policy, and how an auditor can reconstruct what happened. Google’s answer is to make identity, access control and logging native parts of the agent stack rather than optional add-ons.
Google is also wrapping those controls inside a broader governance layer. Agent Registry is intended to act as a central catalogue of approved internal agents, tools and skills, while Agent Gateway and Model Armor are meant to filter traffic, enforce policy and reduce risks such as prompt injection, data leakage and unsafe tool use. In the Gemini Enterprise app itself, companies get a single place where staff can discover, create, share and monitor agents, including long-running agents working through complex tasks over hours or days. That model reflects a growing enterprise preference for visibility and administrative control over decentralised experimentation.
The commercial logic is equally clear. Google is trying to close ground on rivals that have pushed hard into enterprise AI, while also responding to fears of “agent sprawl” inside large organisations. By bundling security, governance, no-code design tools, model infrastructure, third-party connectors and an agent marketplace into one system, the company is offering not just models but an operating framework for AI-led work. That puts it into sharper competition with Microsoft, Amazon, OpenAI and Anthropic for corporate budgets that are increasingly tied to automation rather than experimentation.
Google is trying to strengthen that ecosystem effect with scale and incentives. The company said the Gemini Enterprise app can surface internally built agents alongside third-party agents from providers such as Oracle, Salesforce, ServiceNow, Adobe and Workday, all within the same governed environment. It also announced a $750 million partner fund to speed development and adoption of agentic AI through its partner network, a sign that it wants consultants, software vendors and systems integrators to pull customers deeper into the platform.
