Key Takeaways
- Google rebranded Vertex AI as the Gemini Enterprise Agent Platform at Cloud Next 2026
- A MuleSoft survey of IT leaders found organisations use 12 AI agents on average
- The A2A (Agent-to-Agent) protocol is supported by more than 150 organisations, with production deployments across several industries
- Project Mariner web-browsing agent scores 83.5% on WebVoyager and handles 10 concurrent tasks
- Danfoss automated up to 80% of transactional decisions in email-based order processing with a partner's platform on Google Cloud
The Facts
Google used Cloud Next 2026 to execute one of the most significant rebranding and consolidation moves in its cloud history. Vertex AI — the company's ML platform that has anchored its enterprise AI strategy since 2021 — has been renamed the Gemini Enterprise Agent Platform. The new name reflects a strategic bet: that the next era of cloud computing is defined by agents rather than models, and that Google's competitive position depends on owning the full stack from infrastructure to the productivity applications that enterprise employees use daily.
Disclosure: RECATOOLS is written with the assistance of Claude, made by Anthropic, which is also named in this article. Readers should weigh that interest.
The announcements at Cloud Next were extensive. Workspace Studio introduces a no-code agent builder allowing non-technical employees to create agents that automate workflows within Google Workspace. The Model Garden now includes over 200 models including Anthropic's Claude, giving customers model choice within Google's infrastructure. The Agent Development Kit (ADK) v1.0 released stable versions across four programming languages.
The A2A (Agent-to-Agent) protocol, which reached its first stable version, 1.0, this year, is supported by more than 150 organisations according to the Linux Foundation, with production deployments across several industries. It provides a standardised communication layer for agents from different vendors to collaborate without bespoke integration work. By late 2025, more than 10,000 public MCP servers were deployed — and A2A extends this interoperability further by enabling agents to delegate tasks to other agents rather than just to tools.
Google Cloud customer case studies provide concrete productivity data. Danfoss, the Danish industrial manufacturer, worked with Go Autonomous, a software provider it found through Google Cloud Marketplace, to automate up to 80% of transactional decision-making in email-based order processing, potentially cutting average customer response time from 42 hours to near real time. Suzano, a Brazilian pulp and paper company with 50,000 employees, built a generative AI tool that lets staff query company data in natural language, cutting the time for one materials query from about two minutes to eight seconds.
Technical Deep-Dive
The renaming of Vertex AI to Gemini Enterprise Agent Platform signals a fundamental architectural shift in how Google thinks about enterprise AI infrastructure. The previous Vertex AI model positioned Google as a platform for machine learning workloads — training models, serving inference, managing ML pipelines. The Gemini Enterprise Agent Platform positions Google as the operating environment for AI agents that persist, act, and collaborate across business systems.
Project Mariner, Google DeepMind's web-browsing agent, provides a concrete example of this shift. Scoring 83.5% on the WebVoyager benchmark and handling ten concurrent tasks on cloud-based virtual machines, Mariner automates shopping, information retrieval, and form-filling at a level of reliability that makes it genuinely deployable for enterprise procurement and research workflows — not merely impressive in controlled demos.
The MCP (Model Context Protocol) and A2A protocol combination creates a two-layer interoperability stack. MCP standardises how agents connect to tools (databases, APIs, file systems). A2A standardises how agents communicate with and delegate to other agents. Together, they enable multi-agent workflows where a coordinator agent decomposes a complex task and routes subtasks to specialist agents — each potentially running on different platforms from different vendors.
The ASEAN Perspective
Google Cloud has significant infrastructure presence across ASEAN, with data centres in Singapore, Jakarta, and Kuala Lumpur. The Gemini Enterprise Agent Platform is fully available across these regions, meaning ASEAN enterprises can deploy agent workloads on local infrastructure without routing sensitive data to US data centres.
Singapore has generally led the region in enterprise AI adoption. For Malaysia, Indonesia, and the Philippines, the agent deployment wave is approaching — and the platform announcements at Cloud Next 2026 represent the infrastructure that will underpin it.
The A2A protocol is particularly relevant for ASEAN enterprises operating across multiple countries with different regulatory requirements. An agent serving a Malaysian customer can delegate data-localisation-sensitive tasks to a locally deployed agent while routing other tasks to a Singapore-hosted orchestration layer — all within a single A2A-compliant workflow.
Try our QR Code Generator to create quick-access codes for your enterprise documentation and agent workflow guides.
RECATOOLS Verdict
Google's bet at Cloud Next 2026 is coherent: own the full stack from TPU chip to Gmail inbox, and enterprise customers who adopt AI agents at scale will choose the single-vendor vertically integrated platform over the assembled multi-vendor alternative.
Whether that bet pays off depends on execution. Google Cloud holds approximately 11% of the cloud infrastructure market versus AWS's 31% and Azure's 25%. The agentic era reshuffles some competitive dynamics — particularly favouring platforms that combine AI models with productivity suites — but it does not eliminate the installed base advantages of AWS and Azure.
For ASEAN CIOs, the practical implication is that agent deployment infrastructure is now available from all three major cloud providers, with broadly comparable capabilities. The differentiator is increasingly the quality of the agent workflows built on top of the infrastructure, not the infrastructure itself.
Frequently Asked Questions
It was rebranded as the Gemini Enterprise Agent Platform at Cloud Next 2026, reflecting Google's strategic shift from ML model serving to enterprise agent deployment.
Agent-to-Agent protocol v1.0 — a standardised communication layer that allows AI agents from different vendors to collaborate and delegate tasks to each other.
Google DeepMind's web-browsing agent that scores 83.5% on the WebVoyager benchmark and can handle 10 concurrent browser tasks — available to Google AI Ultra subscribers.
Yes — Google Cloud has data centres in Singapore, Jakarta, and Kuala Lumpur, enabling local data residency for agent workloads.
A no-code agent builder within Google Workspace that lets non-technical employees create agents to automate workflows without writing code.
Correction, 18 October 2026: An earlier version attributed an 89% adoption figure and a 12-agent average to a Google survey; the 12-agent figure comes from a MuleSoft survey. The A2A adoption figure and the Danfoss and Suzano results have also been stated more precisely.
View author profile → · Editorial policy
