The Rebrand That Changed the Conversation
At Cloud Next ’26 in Las Vegas last week, Google did something it has been building toward for three years: it collapsed its fragmented AI product lineup into a single platform and made agentic AI the central pitch to enterprise customers. The headline move was renaming Vertex AI to the Gemini Enterprise Agent Platform — absorbing Agentspace and the enterprise tier of Gemini Code Assist into one unified product with per-agent pricing, a 200-model Model Garden, and a governance layer that had not existed before.
For existing Vertex AI customers, Google was careful to avoid a breaking change. SDKs, billing, and APIs migrate under the Gemini Enterprise namespace without disruption, and the ADK (Agent Development Kit) reached v1.0 stable across four languages: Python, Java, Go, and Node.js. But the buyer persona has visibly shifted. “Vertex AI” sold API access to ML engineers. “Gemini Enterprise” is selling workflow outcomes to operations leaders — a different budget, a different conversation, and a different success metric.
The Model Garden is now genuinely multi-vendor: Anthropic Claude Opus 4.7, Meta Llama 4, Mistral Forge tunings, and DeepSeek V4 sit alongside Gemini 3.1 Pro. Per-call governance and routing are handled centrally by the platform, which addresses a real headache for enterprise compliance teams managing multiple models across business units. Instead of negotiating four separate enterprise contracts and building four separate audit pipelines, organizations get a single control plane.
Workspace Studio: AI Agents for Non-Engineers
The announcement most likely to reach the largest number of employees is Workspace Studio: a no-code agent builder that works across Gmail, Docs, Sheets, Drive, Meet, and Chat. Users describe what they want automated in plain language — “every Friday, ping me to update my tracker” — and Gemini 3.1 constructs the agent. Connections to Asana, Jira, Mailchimp, Salesforce, and external APIs via webhooks ship out of the box, with Apps Script available for custom logic.
The early scale is notable. Google says Workspace Studio has powered more than 20 million automated tasks in the past month, ranging from routine status reminders to more complex workflows like legal notice triage and travel request processing. The platform is rolling out to all Google Workspace Business, Enterprise, and Education customers across both Rapid Release and Scheduled Release domains.
Most employees do not interact with an ADK or a managed runtime — they interact with their inbox. If Workspace Studio delivers on its preview results, it moves agentic AI from the IT roadmap into daily non-technical workflows without requiring a single line of code from the end user. That is a harder problem than building better models, and it is precisely the reason Microsoft’s Copilot gained traction in enterprise accounts that cared nothing about benchmark scores.
There is a governance angle too. Workspace Studio agents operate inside the same AI Control Center and agent management console as developer-built agents, giving IT administrators visibility into what agents exist, which data they access, and what they have executed. For regulated industries, that audit trail is as important as the automation itself.
The A2A Protocol Reaches 150 Organisations in Production
The Agent2Agent (A2A) protocol — Google’s open standard for letting agents from different vendors hand off tasks to each other — shipped version 1.2 at Cloud Next ’26 and is now running in production at 150 organizations. Governance was transferred to the Linux Foundation’s Agentic AI Foundation, a move designed to reassure enterprises that the protocol will not be locked down or quietly deprecated when Google’s strategic priorities shift.
This matters more than it might appear. According to Google’s AI Agent Trends report released at the conference, the average enterprise now runs 12 AI agents, and 89% of business teams say they are already using agents in some form. The most common use cases are customer service (49%), marketing (46%), security operations (46%), and IT support (45%). Agents that cannot interoperate create integration overhead that quickly erodes the productivity gains they are supposed to deliver.
Cloud Next ’26 showcased two production deployments that illustrate what interoperability enables in practice. Danfoss, the Danish industrial manufacturer, automated 80% of transactional decisions in email-based order processing using Gemini agents, cutting average response times from 42 hours to near real-time. Suzano, a Brazilian pulp and paper company, deployed an agent that translates natural language into SQL queries for 50,000 employees — reducing query turnaround time by 95%. Neither deployment is a controlled demo; both run on live business data at scale.
What This Means for Enterprise Buyers
The strategic play at Cloud Next ’26 was clarity. For three years, Google’s enterprise AI story was fragmented: Vertex AI for developers, Gemini for knowledge workers, Agentspace for enterprise search, Code Assist for engineers. Competitors landed simpler narratives. The unified Gemini Enterprise pitch — spanning chip infrastructure through managed runtime to the Workspace inbox — is a deliberate response to Microsoft Copilot’s end-to-end positioning and OpenAI’s enterprise API offerings.
The risk is the familiar one: platform consolidation announcements are easier to make than to sustain. Per-agent pricing is conceptually cleaner than per-token billing, but organizations running a dozen or more agents will still need careful cost modelling before committing at scale. And while handing A2A governance to the Linux Foundation is a positive signal, the protocol’s real test is how gracefully it handles edge cases — when a Gemini agent needs to hand off a partially completed task to a ServiceNow or Salesforce agent mid-workflow, with state preserved and permissions validated at each hop.
As we have tracked in our analysis of why only 5% of enterprises see measurable AI ROI in 2026, the bottleneck is rarely model quality. It is integration depth and organizational change management. Workspace Studio targets the change management problem directly, meeting workers in the tools they already use. The A2A protocol targets the integration problem, though evidence from 150 production deployments is encouraging rather than conclusive.
The practical evaluation question for buyers is not whether Gemini Enterprise is technically capable — it clearly is. It is whether operating agents inside a single governed platform is measurably easier to manage, audit, and cost-control than running equivalent agents across Vertex AI, OpenAI, and Anthropic APIs separately. For organizations already committed to Google Workspace, the answer is probably yes. For those with a heterogeneous stack, the evidence will take another quarter or two to arrive.
Further Reading
- Introducing Gemini Enterprise Agent Platform — Google’s official platform overview covering ADK v1.0, Model Garden details, and the per-agent pricing model.
- Google Cloud Next 2026: AI agents, A2A protocol, Workspace Studio, and the full-stack bet — The Next Web’s comprehensive three-day recap placing the announcements in competitive context.
- Google Cloud Next 2026: The End Of The AI Pilot Era — Forrester’s analyst take on why Google’s consolidation signals a structural shift in enterprise AI procurement.