Corporate Business Alliance

CBA-GAI · Study material

Contents

The complete contents of the CBA Certified Generative AI Practitioner study material: every chapter and lesson, the reading time of each, and where the assessed workbooks fall. One lesson is open to read in full.

Lessons

81

Reading time

5 hours

Workbooks

19

Words

59,989

These figures are counted from the material itself. The reading time assumes a steady pace and no re-reading, and does not include the time spent building the workbooks.

Free, no account

Read a complete lesson

Why the same prompt gives a different answer twice, from Generative Model Foundations, as a candidate reads it, slides included.

Read the lesson

Contents

7 chapters, 81 lessons. Lessons marked workbook carry a spreadsheet you build and submit, marked against the published tolerance. The award requires 4 of the 19 completed to that tolerance.

Introduction

1 lesson · 2 min

  1. CBA-GAI Study Guide2 min

01Generative Model Foundations

13 lessons · 45 min

  1. What this chapter covers2 min
  2. What the model is actually doing when it produces text5 min
  3. Tokens: the unit the model actually works inworkbook3 min
  4. The context window, and why a long document behaves differently from a short one4 min
  5. Training, fine-tuning and in-context learning are three different things4 min
  6. Why the same prompt gives a different answer twiceFreeworkbook4 min
  7. Fluent wrongness: what models invent, and whenworkbook5 min
  8. Knowledge cutoffs, and the slope before them2 min
  9. What the model does not know about itself3 min
  10. Diffusion models for images, at the same conceptual level3 min
  11. Capability and reliable capability, which is the distinction this credential rests on5 min
  12. Common pitfalls3 min
  13. Chapter summary2 min

02Prompt Design and Iteration

13 lessons · 61 min

  1. What this chapter covers2 min
  2. The prompt as a specification, not a request3 min
  3. The six components that reliably matter8 min
  4. Being specific about what you do not want3 min
  5. Why examples outperform description, and how many are enough4 min
  6. Decomposition: one prompt asking for five thingsworkbook4 min
  7. Giving the model a way to say it does not knowworkbook6 min
  8. Iteration as a disciplineworkbook5 min
  9. A pattern library for the work Lindenmoor actually doesworkbook15 min
  10. Portability, and why a prompt library is a versioned thing3 min
  11. The limits: what no prompt can fix4 min
  12. Common pitfalls2 min
  13. Chapter summary2 min

03Retrieval and Grounded Workflows

15 lessons · 48 min

  1. What this chapter covers2 min
  2. Grounding: giving the model the source instead of trusting its memory3 min
  3. Retrieval fixes accuracy problems, and no capability problems3 min
  4. How retrieval works, and what each step can get wrong3 min
  5. Chunk size and boundaries change the answer3 min
  6. The two ceilings: perfect retrieval, and what the corpus actually containsworkbook3 min
  7. Measuring retrieval separately from generationworkbook3 min
  8. Citation makes an answer into a different product5 min
  9. Grounding the marketing blurb, where the source was never in the corpus4 min
  10. The four ways retrieval failsworkbook4 min
  11. Retrieval inherits the corpus's permissions, or it destroys them2 min
  12. Choosing between prompting, retrieval and fine-tuning5 min
  13. Operating a grounded workflow: refresh, abstention and sampling4 min
  14. Common pitfalls2 min
  15. Chapter summary2 min

04Multimodal Creation and Editing

11 lessons · 27 min

  1. What this chapter covers2 min
  2. What an image model is doing, and what follows from it3 min
  3. Prompting for images: five things, in order2 min
  4. Iterating on an image rather than regenerating it2 min
  5. Editing an existing image: four operations, four different risks3 min
  6. Reading an image: extraction, description, and the 40,000-asset problemworkbook5 min
  7. What a bad alt text costs, and why review is not optional2 min
  8. Provenance, watermarking and disclosure2 min
  9. Rights: generating, owning, and alteringworkbook2 min
  10. Common pitfalls2 min
  11. Chapter summary2 min

05Output Evaluation and Quality Control

12 lessons · 45 min

  1. What this chapter covers2 min
  2. Why generative output cannot be scored like a classifier3 min
  3. Building a rubric two reviewers can actually apply4 min
  4. Measuring agreement, and reading disagreement as a rubric faultworkbook4 min
  5. Sampling: how many, chosen how, and why the suspicious ones tell you nothing4 min
  6. Golden sets and regression checking4 min
  7. Automatic checks that genuinely work6 min
  8. Designing human review so that it does not become signingworkbook4 min
  9. Three different problems: marketing copy, manuscript summaries, alt textworkbook5 min
  10. Deciding good enough, and who decides5 min
  11. Common pitfalls2 min
  12. Chapter summary2 min

06Safe Use, Data Care and Work Integration

16 lessons · 47 min

  1. What this chapter covers2 min
  2. The question is what the contract says, not what the interface looks like3 min
  3. Retention, training and how to find out what a provider actually doesworkbook3 min
  4. Personal data: a name in a prompt and a name in a training set are different problems5 min
  5. Confidential third-party material: the manuscripts nobody may paste anywhere4 min
  6. Intellectual property: three separate questions people merge into one4 min
  7. Attribution and disclosure: who has to be told, and what telling them means2 min
  8. Accuracy obligations, and where the liability actually sits3 min
  9. Prompt injection: instructions arriving inside the contentworkbook3 min
  10. Choosing which tasks to delegateworkbook3 min
  11. Redesign the process, do not bolt the tool onto it2 min
  12. Measuring whether it helped, honestlyworkbook4 min
  13. Skill atrophy, and the junior who never learns to draft2 min
  14. A policy people follow3 min
  15. Common pitfalls2 min
  16. Chapter summary2 min

What the material is

The study material is written to the discipline’s domains and learning objectives, chapter by chapter, with worked examples and unscored checks in the lessons and assessed workbooks where the level requires applied work. It does not contain a bank of practice questions matching the examination; the specimen paper shows the style and standard of the examination, with a written rationale for every option.

Back to CBA-GAICertification scheme