Agentic AI with Andrew Ng

Building Live Voice Agents with Google’s ADK — A Practical Course by DeepLearning.AI
I present a concise briefing on a new DeepLearning.AI course about agentic AI.
This course focuses on building live voice agents and agentic workflows.
Why does this matter to you?

• Because agentic AI can automate multi-step business processes that single prompts cannot.
• Because the course teaches practical implementation in Python from first principles.
• Because the instructor is Andrew Ng, and the course awards a certificate after completion.

Agentic Design Patterns
Agentic AI is about iterative, multi-step workflows rather than single-shot responses.
It uses patterns like reflection, tool use, planning, and multi-agent coordination.
What do these patterns enable?
They let models plan, execute, and improve over several steps.
I recently prototyped a small planning agent to coordinate data collection and found iteration reduced errors quickly.
Example prompt snippet for a reflection loop:

System: You will review the draft report and list three improvements. 
User: Draft report attached. 
Assistant: Step 1 - Check facts. Step 2 - Suggest structure changes. Step 3 - Propose visualizations.

Tool Integration and Execution
Tool use means connecting models to databases, APIs, web search, or code execution.
Why connect tools?
Because real tasks need world access and reliable outputs.
The course covers function calling, code execution, and the Model Context Protocol (MCP).
Example code prompt for calling a function:

Request: fetch_latest_prices(product_id=12345) 
Assistant: Call function and return JSON of prices by region.

Evaluation, Planning and Multi-Agent Systems
Good systems need metrics, error analysis, and production readiness.
What makes a planning agent different?
It composes multi-step plans and can delegate to specialized agents.
The course includes a capstone to build a full research agent.
Example planning prompt:

Task: Create a 4-step plan to research competitor features and summarize differences.

Wrapping up
This course teaches practical, production-oriented agentic AI skills.
You will learn four core design patterns and how to evaluate systems in the wild.
It is aimed at developers with intermediate Python and basic LLM experience.
Interested in the syllabus or enrollment details?
Find the course here: https://www.deeplearning.ai/courses/agentic-ai/?utm_campaign=24113415-agentic-ai&utm_content=350623501&utm_medium=social&utm_source=twitter&hss_channel=tw-992153930095251456
For C-level and technical leaders, agentic workflows change what automation can do.
They make complex, multi-step processes automatable and measurable.
This is where generative AI delivers tangible business value through improved efficiency and repeatable outcomes.