Google AI product names in 2026: the Vertex AI, Gemini Enterprise, and Agent Studio rebrand glossary

If you are studying for the Google Cloud Generative AI Leader certification, the single biggest source of lost points is not the concepts — it is the product names. Google renamed much of its enterprise AI stack across late 2025 and April 2026, and the official exam guide predates parts of that rebrand. So you will see both old and new names, sometimes in the same question set.

This is a living reference. Learn each mapping in both directions — old-to-new and new-to-old — because the exam may present either.

The core renames

Vertex AI → Gemini Enterprise Agent Platform

The platform where you build, tune, and deploy models on Google Cloud. Same platform, new name. Model Garden — the catalog of first-party and open models — persists inside it. When study material says "Vertex AI," read it as "the Gemini Enterprise Agent Platform."

Agentspace → Gemini Enterprise

The umbrella product. Gemini Enterprise combines:

• an AI assistant for employees,
• an agentic platform for building and running agents, and
• enterprise search with connectors to sources like SharePoint, Confluence, Jira, and ServiceNow.

If a question describes "org-wide, connector-based enterprise search," the answer is Gemini Enterprise (via Agent Search), not a point solution.

Agent Search

The multi-turn, multimodal enterprise search capability inside Gemini Enterprise. This is the "search across all our internal docs in natural language" answer.

Model Garden

The model catalog — first-party models (Gemini, Imagen, Veo) plus open models (like Gemma) and select third-party models. It survived the rebrand as a sub-component of the Gemini Enterprise Agent Platform.

The trap pairs

Google AI Studio vs Agent Studio

This is the most common naming trap.

If a question mentions enterprise governance, controls, or oversight, the answer is Agent Studio. If it mentions individual prototyping or the Gemini Developer API, it is Google AI Studio.

Conversational Agents (formerly Dialogflow CX)

Conversational Agents is the current name for Dialogflow CX. The rebrand is still in transition, so expect both names. Two things to remember:

1. It lives in the Customer Engagement Suite (alongside Agent Assist, Conversational Insights, and Contact Center as a Service).
2. It is hybrid — deterministic flows for known intents plus generative responses for open-ended queries. It is not "just a chatbot," and it is not purely generative.

Gemini Enterprise vs NotebookLM Enterprise

These are complementary, not competitors:

• Gemini Enterprise — broad, org-wide search and agents.
• NotebookLM Enterprise — curated deep research over a chosen corpus, running inside the customer's own Cloud project.

A question that pits them against each other is testing whether you know they solve different problems.

Two honest caveats

1. Google's own docs may lag. During a rebrand, console labels, help pages, and the exam guide do not all update at once. Do not panic if you see "Vertex AI" in an official-looking source — recognize the platform behind it.
2. Don't over-specify the hardware. Google describes its accelerators generically as custom-designed TPUs. For a business-level exam, "custom-designed TPUs, GPUs, and global data centers" is the safe framing — avoid guessing a specific TPU generation number.

Quick reference table

Update — September 2, 2026: The naming sprawl isn't limited to the enterprise stack — Google's consumer/Workspace image tools hit the same wall with the launch of Google Pics, a new standalone editor that sits alongside Nano Banana, Imagen, and the Gemini app's own image generation. See that post for a table untangling what each one actually is.

How should you use the glossary when documentation disagrees?

Record the exact term, the page where it appears, and the capability described around it. A rename can change navigation and packaging without changing every API identifier. If a tutorial uses an older name, determine whether the underlying task still matches the current product rather than replacing words mechanically.

For a hands-on task, prioritize the current official documentation for the specific service and interface you are using. For exam study, keep the terminology in the current official exam guide visible as well. The two can differ during a transition. Learn the capability that connects the names so either wording remains understandable.

This glossary provides a historical mapping, not a guarantee that every legacy SKU, console page, or API endpoint has moved identically. Google's Gemini Enterprise overview is a starting point for current product boundaries. Follow its links to the exact service before making an implementation decision.

A worked scenario: employee search versus application development

Suppose a company wants employees to search approved documents across connected business systems. Start with the users, sources, and access rules. The central requirement is that an employee receives only information they are authorized to see, with evidence they can inspect. A product name containing “agent” does not by itself answer those requirements.

Now consider a developer building a customer-facing application that calls a model API. That workflow needs API credentials, an integration contract, deployment decisions, and application-level evaluation. It is a different task from enabling a workforce assistant, even if both involve Gemini models. Keep the model family and the surrounding product distinct in your notes.

Finally, consider a researcher collecting a small set of approved documents for a focused analysis. Curating that corpus is different from connecting an organization-wide search system. The right question is which workflow fits the task and its governance requirements, rather than which product name sounds most comprehensive.

Which distinctions remain useful after the next rename?

Separate five layers: model, developer API, development environment, application or agent platform, and end-user experience. A model generates or interprets content. An API exposes capabilities programmatically. A studio helps a person experiment or build. A platform supplies deployment and operational controls. An end-user product packages the capabilities for a particular audience.

Names can span more than one layer, so this is a reading aid rather than a rigid taxonomy. When a headline says a new product is “powered by Gemini,” ask which layer changed. A new employee interface does not necessarily imply a new model, and a new model release does not imply the employee product immediately offers every feature.

Keep governance questions explicit. Where does data go? Which identity controls access? Which administrator owns the environment? What gets logged? Which tools can an agent invoke? These are architecture questions that need service-specific answers. Describing a prototyping surface as having “no governance” is too absolute; inspect its available controls and the account or cloud context in which it runs.

How can students practice the naming distinctions?

Write three short scenario cards without vendor names: a team needs grounded answers over internal documents; a developer needs to test prompts with a model API; a contact center needs predictable flows plus generative responses. For each card, identify the needed capability and then locate the relevant product family.

Swap the legacy and current labels on the cards and repeat. If the answer changes merely because the label changed, revisit the capability description. The point is to understand what the service does, not memorize a chain of arrows without recognizing the underlying task.

Add a fourth card where the correct answer is insufficient information. For example, “build an enterprise agent” omits sources, allowed actions, users, and deployment constraints. Explain which questions must be answered before choosing a service. This prepares you for real architecture conversations as well as multiple-choice questions.

How should a team maintain its own naming reference?

Keep a short internal table with the legacy term, current documentation entry, task it supports, and date checked. Link directly to primary documentation instead of copying long feature descriptions that become stale. Add API or command identifiers separately when they remain unchanged.

Assign an owner to update the reference when a working integration or runbook changes. A quarterly review may be sufficient for background terminology, while a deployment task deserves a fresh check. Treat the glossary as navigation infrastructure: it helps people find the right contract, but it does not replace the contract itself.

When publishing your own tutorials, retain the old term in an explanatory note if readers still search for it. State the current term near the setup instructions, and name the date of the documentation check. That keeps a guide discoverable while avoiding the impression that an old screenshot establishes the current console layout.

Related on explainx.ai

• Google Pics: Workspace's new AI image editor explained — untangling Pics vs Nano Banana vs Imagen
• Google Cloud Generative AI Leader certification guide — the full exam breakdown
• Certification study guide — domains, scenarios, task map
• Learning pathway — articles mapped to exam domains
• Google Cloud Next 2026: TPU and Gemini Enterprise Agent Platform — the rebrand in context

Product names are summarized from Google Cloud's public messaging as of mid-2026 and change frequently; verify current naming on Google Cloud before relying on it. explainx.ai is not affiliated with Google Cloud.