What to Read
Agents are not just chatbots. They can observe, interpret, plan, act, and use tools to
complete goal-oriented workflows.
Exam shortcut: if the question needs live inventory, a database lookup, custom business logic,
or multi-step action, think agent plus tools.
Agent Reasoning Loop
- Observe: gather input, context, and state.
- Interpret: understand the user's goal and constraints.
- Plan: decide which step or tool should come next.
- Act: call a tool, query data, produce output, or trigger a workflow.
- Review: continue, escalate, or stop when the goal is reached.
Google Cloud Tooling
| Service | Agent use case |
|---|
| Cloud Storage | Store and retrieve files, documents, media, or intermediate artifacts. |
| Cloud SQL / Spanner / Firestore | Query structured or operational data. |
| Cloud Run Functions | Execute custom business logic as a tool. |
| Document AI | Extract structured information from unstructured documents. |
| Maps APIs | Location, distance, routing, and geocoding tasks. |
| Vertex AI | Build models and agents that can be used by other agents. |
Product Distinctions
| Product | Best fit |
|---|
| Gemini | General multimodal generation, productivity, and ideation. |
| NotebookLM | Focused research over uploaded sources. |
| Gemini Enterprise | Enterprise-wide assistants, custom agents, and multimodal search. |
| Vertex AI Search | Ground answers in first-party, third-party, or Google knowledge sources. |
| Vertex AI Agent Builder | Build managed custom agents with tools and enterprise data. |
