Agentic AI

What the course covers

The course has 31 video lessons, 7 code examples and 8 graded assignments across five modules, about 9 hours 55 minutes in total. The topics:

• Reflection: having a model critique and improve its own output, and when this measurably helps.
• Tool use: letting a model call functions, APIs, databases and code execution.
• Evaluation and error analysis: building evaluations for each step of an agentic workflow and finding which component is causing failures. This is the part most tutorials skip, and one of the most useful here.
• Planning: letting the model decide a sequence of steps.
• Multi-agent collaboration: splitting work across several specialised agents and coordinating them.

The course is taught in Python without committing you to a particular agent framework, so the ideas carry across tools.

Free versus Pro

You can audit the course for free, which gives you the videos and the forum. The quizzes, graded assignments and certificate need a DeepLearning.AI Pro membership, listed at $25 a month billed annually or $30 month to month as of September 2026. If you learn by building, the assignments are where the value is, so consider paying for a month and finishing the course in that time.

Before you start

The course asks for intermediate Python and a basic understanding of large language models. You should be comfortable calling an LLM API and reading JSON. If you are not, a short introductory course on prompting or the API first will make this much easier.

How to get the most from it

• Do the assignments rather than only watching. The patterns are simple to describe and harder to make reliable.
• After each module, apply the pattern to a small project of your own, such as a research assistant or a support bot that looks things up.
• Take the evaluation module seriously. Being able to say which step of your agent fails and how often is what separates a demo from a system.
• Plan on about two weeks at five hours a week.

Where it fits

• Before it: a basic prompting or LLM API course.
• Alongside it: Anthropic's article Building Effective Agents, which argues for starting with the simplest workflow, is a good companion read.
• After it: a retrieval course (DeepLearning.AI's RAG course) and the documentation for whichever agent framework or SDK you choose to build with.

Common questions

Is it only videos? No. There are code examples and graded programming assignments, although the graded parts need Pro.

Does it teach a specific framework such as LangChain? It focuses on the patterns in Python, so you can apply them in any framework or with plain API calls.

Do I get a certificate? Yes, with Pro, after completing the graded work.

Best for: Python developers who want a structured, evaluation-driven way to design and build AI agents.