1.2 From Vertex AI to Gemini Enterprise Agent Platform: The Product Name Map

If you learned Google Cloud ML before 2026, many product names on the exam will look unfamiliar. On April 22, 2026, Google's release notes announced that Vertex AI features now live in Gemini Enterprise Agent Platform, usually shortened to Agent Platform. Google described it as the evolution of Vertex AI: all Vertex AI services and future roadmap work are delivered through Agent Platform rather than a standalone service. The PMLE exam guide dated June 1, 2026 uses these new names throughout. This guide does too, with the former name in parentheses where that helps.

Core ML Name Changes

What did not change

• The REST endpoint is still aiplatform.googleapis.com. For example, a Gemini endpoint URL still looks like https://us-central1-aiplatform.googleapis.com/v1/projects/.../publishers/google/models/....
• The Python SDK package is still google-cloud-aiplatform, so code snippets on the exam may still import aiplatform.
• Bookmarked console links redirect to the new Agent Platform navigation.
• Features kept working. The rename is about branding and navigation, not a migration of your models or pipelines.

Two products named "Gemini Enterprise"

Gemini Enterprise is the workplace app that employees use to find information and run agents. Gemini Enterprise Agent Platform is the developer platform for models, training, tuning, serving, and agents. The PMLE exam is about the second one.

Data and Analytics Stack Renames

The exam page also mentions updates to Google Cloud's data and analytics stack. The renames most likely to appear in ML scenarios are:

Deprecations That Change the Right Answer

Renames don't change correct answers, but deprecations do. Know these:

1. Vertex Explainable AI was deprecated on March 16, 2026. No new features are being added, and the APIs are scheduled to shut down on March 16, 2027. Google suggests open-source libraries such as SHAP and LIME as alternatives. The exam still lists model explainability, so learn both the built-in attribution methods and the open-source path (Chapter 18).
2. Feature Store (Legacy) and Feature Store Optimized online serving were deprecated on February 17, 2026, with shutdown on February 17, 2027. The current Feature Store uses BigQuery as the offline store and serves online through Bigtable online serving. Google directs embedding workloads to Vector Search (Chapters 5 and 14).
3. AutoML Text and AutoML Video are retired. Google now directs text classification, entity extraction, sentiment, and video use cases to Gemini prompting and supervised tuning. Agent Platform AutoML now covers image data (classification and object detection) and tabular data (classification, regression, and forecasting).
4. Workbench managed notebooks and user-managed notebooks reached end of life on March 30, 2026. Workbench instances and Colab Enterprise are the supported notebook environments (Chapter 6).

How to Read Older Practice Material

Much third-party material still says "Vertex AI." Translate names as you read, but check any design that depends on a retired capability. For example, a legacy practice answer might recommend AutoML Text for sentiment classification. On the current platform, the best low-code answer is a Gemini model called through Agent Platform, with supervised tuning if prompting falls short.

Use this three-step check whenever an older explanation names a product:

1. Translate the name. "Vertex AI Pipelines" becomes Agent Platform Pipelines, and "Vertex AI Endpoints" becomes Agent Platform Endpoints. The idea behind the answer usually holds.
2. Check the capability's status. If the answer needs AutoML Text, AutoML Video, legacy Feature Store entity types, managed notebooks, or long-term Explainable AI, look for the current replacement.
3. Re-apply the constraint. The current exam favors Google Cloud-native managed services. If a scenario asks for the least operational overhead, a managed Agent Platform or BigQuery capability usually beats a self-managed cluster running the same open-source tool.

Names that did not change

Some products you will see in ML scenarios kept their names: BigQuery, BigQuery ML, Cloud Storage, Dataflow, Pub/Sub, Cloud Run, Google Kubernetes Engine (GKE), Cloud Build, Artifact Registry, Cloud Logging, Cloud Monitoring, and Model Armor. Gemini model families (Pro, Flash, and Flash-Lite), Imagen, and Veo are model names, not platform names. Their version numbers change often, so exam questions focus on choosing a model tier rather than memorizing version strings.