SAP AI Foundations

Go from zero AI knowledge to confidently designing, building and governing SAP Joule agents — a progressive, hands-on series.

1. 1
   Video

   AI, Machine Learning & Generative AI: The Map

   Lesson 1 — A client just asked you to explain the difference between AI, ML and Generative AI in one sentence. Here's the map that makes you sound like you've known this for years — plus the one property of generative AI that will bite you in your very next demo if you don't understand it.

2. 2
   Video

   Inside a Large Language Model: Tokens, Context & Cost

   Lesson 2 — Why did that Generative AI Hub call cost three times more than yesterday's? Tokens, context windows, and inference explained the way you actually need them: as an engineer sizing a bill, not a data scientist writing a paper.

3. 3
   Video

   Prompting That Works (+ Context Engineering)

   Lesson 3 — Two consultants ask an LLM the exact same question. One gets a usable answer in one shot. The other gets three rounds of back-and-forth and a worse result. The difference isn't the model — it's a skill you can learn in twelve minutes.

4. 4
   Video

   Embeddings & Semantic Search

   Lesson 4 — A user searches your knowledge base for 'PO' and gets nothing, even though the exact answer sits in a document that only ever says 'purchase order.' Here's the fix that's been hiding in plain sight, and why it's the real foundation under everything SAP calls 'grounding.'

5. 5
   Video

   Retrieval-Augmented Generation (RAG)

   Lesson 5 — Your client asks Joule about a custom Z-table, and it answers instantly, fluently, and completely wrong, because it was never trained on your system. This is the pattern that fixes that — and the honest limits of what it can and can't do.

6. 6
   Video

   Fine-Tuning vs RAG vs Prompting: Choosing the Right Technique

   Lesson 6 — Someone on your project is about to propose 'let's just fine-tune a model on our data.' It's the single most expensive wrong answer in enterprise AI, and this lesson is the two-minute argument that saves the budget.

7. 7
   Video

   From Chatbot to Agent: Tools, Function Calling & the Reason–Act Loop

   Lesson 7 — There's one sentence about AI agents almost everyone gets backwards, and it's the exact sentence that decides whether your agent design is safe to put in front of a client. Tool calling, structured output, and who's really pulling the trigger.

8. 8
   Video

   The SAP AI Stack: AI Foundation, Generative AI Hub & Joule

   Lesson 8 — Five SAP AI product names, five completely different jobs, and one confusion that quietly costs projects real budget: buying or building the wrong layer because two names sound alike. Here's the map that ends that.

9. 9
   Video

   How SAP Joule Works: Grounding, Orchestration & Agents

   Lesson 9 — 'Joule just knows things.' It doesn't — and the moment you can explain the four-step lifecycle underneath that illusion is the moment you go from Joule user to someone who can actually design for it.

10. 10
    Video

    MCP & A2A: The Protocols That Connect Agents

    Lesson 10 — The difference between building one clever agent and building an agent ecosystem comes down to two acronyms and one design decision most teams get wrong on their first try: reaching for multiple agents before they've earned the complexity.

11. 11
    Video

    Build Your First SAP AI Agent: Joule Studio + MCP

    Lesson 11 — Everything in this series so far was preparation. This is where you actually assemble a working SAP AI agent — and the one discipline (start read-only) that separates a safe first project from a very uncomfortable incident review.

12. 12
    Video

    Evaluating, Testing & Observing AI Agents

    Lesson 12 — Your agent worked perfectly in the demo. Three weeks later, after a quiet model update nobody told you about, it doesn't. This lesson is the discipline that catches that before your client does — and it's a discipline you already know by another name.

13. 13
    Video

    Securing, Governing & Shipping AI for Clean Core

    Lesson 13, the capstone — An agent with write access and no governance is a live incident waiting for a Tuesday. This is the lesson that turns 'it works in the demo' into 'I would stake my professional reputation on this running in production' — and the through-line for the entire series.