Many AI tools can now answer questions about your own files: "Ask your documents", "Chat with your PDF", or a help bot on a website. Most of these tools use a method called RAG. In this short course, you learn how RAG works, what can go wrong, and how to check the answers.
You follow the organizer of a fictional community event, AI for Everyone. The organizer wants an assistant that answers attendee questions from the event documents. Each lesson shows examples of questions, found text, and answers. Then you try it yourself. We wrote the examples for teaching. A real tool may behave in a different way.
Your 25-minute route
| Chapter | What you practice | Lesson time |
|---|---|---|
| 1. Why AI needs your documents | Decide when a question needs RAG | 4 minutes |
| 2. How the search works | Follow a question from search to answer | 5 minutes |
| 3. When RAG goes wrong | Find the reason for a wrong answer | 5 minutes |
| 4. Check the answer | Check sources and test a RAG tool | 4 minutes |
Plan about 18 minutes for the lessons and exercises, 4 minutes for the chapter checks, and 3 minutes for the final test. These times are estimates. Take more time if you need it.
Each chapter has one lesson and two questions. The final test has four new questions, one for each chapter. To pass a chapter check or the final test, you must answer every question correctly. After you submit, you see an explanation for each question. For multiple-choice questions, select all correct answers.
How to practice
You do not need any setup or paid tools. You can do all exercises on paper. You do not need to know how to code.
Generative AI Essentials and Practical Prompting are good courses to take first, but you do not need them. Remember: RAG helps a model use the right information. You still need to check the answer.
What you will learn
- Explain what RAG is and why a model needs it to answer questions about your documents
- Describe how documents are split into chunks and how search by meaning finds them
- Recognize common reasons why a RAG answer is wrong, and name a fix for each
- Check a RAG answer against its sources, and test a RAG tool with a few questions
Before you start
- You know how to use everyday web applications
- You do not need programming experience, an AI account, or paid tools
Course modules
Why AI needs your documents
Learn why a model does not know your documents, and how RAG gives it the right information at the moment you ask.
How the search works
Follow a question through chunks, search by meaning, and the final prompt.
When RAG goes wrong
Learn five common reasons why a RAG answer is wrong, and a fix for each one.
Check the answer
Check a RAG answer against its sources, and test a RAG tool with a small set of questions before you share it.