Gemini Enterprise Agent Platform lets you build, govern and optimize your agents.

Source: Google Date: 2026-04-22 URL: https://blog.google/innovation-and-ai/infrastructure-and-cloud/google-cloud/gemini-enterprise-agent-platform/

Summary

Google launched the Gemini Enterprise Agent Platform at Cloud Next ‘26, combining Vertex AI’s model capabilities with integrated tooling for agent development, security, and DevOps. The platform provides access to Gemini 3.1 Pro, 3.1 Flash Image, and Lyria 3, alongside Anthropic Claude variants, and connects to the Gemini Enterprise app as an organization-wide AI interface. It’s positioned as a unified build-govern-operate layer for enterprise agent deployment.

Implications

The inclusion of Anthropic Claude variants inside Google’s enterprise platform is the most strategically interesting detail in this launch. Google is positioning itself as the neutral orchestration layer — “we support all the models, including our competitor’s” — which is the same move Microsoft made with Azure OpenAI integrations before it became the dominant enterprise AI cloud. It’s a pragmatic bet that enterprises want the platform, not any single model.

This directly challenges AWS Bedrock’s multi-model story and Azure AI Foundry’s agent orchestration push. The differentiators Google is betting on: TPU 8i inference cost advantage, Wiz-backed security governance, and the Gemini Enterprise app as an employee-facing layer that routes into the same platform developers build on.

Three threads:

• Cloud agent platform: This is the product that makes Cloud Next ‘26 coherent. Every other launch (TPU 8i, Gemini Embedding 2, Agentic Data Cloud) is infrastructure under this roof.
• Enterprise positioning vs. OpenAI/Anthropic: By hosting Claude, Google neutralizes the “but Anthropic is safer” objection in enterprise deals while keeping spending inside Google Cloud.
• Governance gap: The build-govern-optimize framing targets the CIO/CISO audience that OpenAI’s enterprise tier hasn’t fully addressed with observable, auditable agent behavior.

Watch:

• Pricing model — is it consumption-based, seat-based, or bundled into Google Workspace Enterprise?
• Whether the Claude integration is a reseller arrangement or a direct API passthrough
• How Agent Studio (low-code) and Agent Designer (no-code) compete with Salesforce Agentforce’s similar tiered builder model