Google expands Gemini Enterprise with governance features

Google has expanded Gemini Enterprise with new agent management, collaboration and governance features. The update adds support for long-running agents and introduces a new agent platform for business users.

The changes target organisations that want AI systems to handle multi-step workflows over extended periods rather than answer one-off questions. The package combines updates to the Gemini Enterprise app with a newly introduced Gemini Enterprise Agent Platform.

Key additions include tools for building agents through natural language prompts or a visual interface, a shared workspace called Projects, an integrated editor called Canvas and a central Inbox for monitoring agent activity. Google also introduced governance features including Agent Identity, Agent Registry and Agent Gateway.

The expansion reflects growing business demand for autonomous software agents that can carry out tasks across multiple systems while remaining subject to internal controls. Customers are looking for systems that can be traced, monitored and managed in regulated environments.

Longer Tasks

A central part of the update is support for long-running agents. These agents can manage multi-step tasks for hours or days, including financial reconciliation and sales prospect sequencing, without constant human prompting.

They are managed through the new Inbox interface, which groups notifications into categories such as items needing user input, errors and completed work. The feature is intended to give teams a single view of ongoing agent activity.

Google has also upgraded Agent Designer, which lets users create agents either with natural language instructions or through a visual workflow builder. The tool combines generative AI with deterministic steps so businesses can inspect, test and approve workflows before deployment.

Human review remains part of the process. Teams can add checkpoints where staff approve actions or provide guidance during a workflow.

Shared Workspace

On the collaboration side, Google introduced Projects, a shared environment where staff and software agents can work together on a common task. The workspace can draw context from sources including Google Workspace, Microsoft OneDrive and team chats.

Canvas, an editor built into Gemini Enterprise, lets teams create and edit documents and slide decks in a single interface with live co-editing. It also supports exporting content to Microsoft Office formats.

Google is also expanding how agents can interact with users. Support for the Agent-to-UI protocol, or A2UI, allows custom agents to generate interactive interface elements inside Gemini Enterprise, including data visualisations and forms.

For technical teams, engineering and IT departments can build coded agents and publish them directly into the Gemini Enterprise app for wider business use. That sits alongside no-code options for non-technical users.

Data And Research

Google is adding two prebuilt agents to the product. The Data Insights agent is designed to analyse both structured data from databases and warehouses and unstructured material such as documents, email and chat. It can generate SQL and search queries and return visualisations based on the results.

An updated Deep Research agent is intended for longer research tasks. It can combine information from the open web and internal company data, working in the background for hours to produce reports and summaries with citations.

Ecosystem Push

Google is also positioning Gemini Enterprise as a broader distribution point for third-party and internally developed agents. It has launched an Agent Marketplace within the Agent Gallery, where users can browse and request external agents from partners including Accenture, Oracle and ServiceNow.

In addition, the company introduced support for Bring Your Own Model Context Protocol integrations. The feature allows administrators to connect Gemini Enterprise to custom or third-party tools, internal workflows and company-built applications.

Gemini Enterprise can work across data held in Google Workspace, Microsoft 365 and other business applications while respecting existing permissions. That approach is intended to address a common business requirement: keeping access controls consistent when AI tools pull information from multiple systems.

Governance Layer

Governance is a major part of the update. Agent Identity assigns digital IDs to agents, Agent Registry provides a catalogue of approved agents, and Agent Gateway acts as a control point for network policies, data access and security rules.

According to Google, Agent Gateway is designed to help organisations prevent agents from sending data to unapproved endpoints and reduce risks such as prompt injection. These controls are included in the Gemini Enterprise portfolio rather than sold as an extra layer.

The new features will roll out to customers over the coming months.