Google: Optimize Agent Behavior

Overview

Turn a basic AI agent into a sophisticated assistant using advanced instructions, model selection, planning capabilities, and structured output. You’ve built your first agent—now it’s time to take it further.

In this course, you’ll advance your skills by learning how to turn a basic AI agent into a sophisticated, precise assistant—applying advanced instructions, model selection, planning capabilities, and structured output patterns. Join the community forum for questions and discussions

Syllabus

• Advanced instruction writing
This module transforms basic instructions into professional agent guidance systems. You’ll learn structured patterns for defining agent persona, role, boundaries, and communication style—turning simple prompts into sophisticated behavioral frameworks grounded in ADK best practices.
• Structured output
This module transforms unpredictable text responses into reliable, structured JSON output that systems can parse and use. You’ll learn how to use Pydantic schemas to enforce output format, making your agents production-ready for integration with applications, databases, and workflows.
• Choosing and configuring models
This module empowers you to strategically select models and configure generation parameters for optimal performance. You’ll learn when to use Flash versus Pro models, how to set safety thresholds, and how to fine-tune temperature and token limits—transforming basic agents into production-optimized systems.
• Planning for complex tasks
This module adds multi-step reasoning to your agents, transforming them from reactive responders into thoughtful problem-solvers. You’ll learn how to enable Gemini’s built-in thinking capabilities using BuiltInPlanner, allowing agents to plan their approach before executing—essential for complex, multi-step tasks.
• Reading List
Reading List

---

Taught by

Google Cloud

---

Subjects

Artificial Intelligence