AI Literacy Foundation Course

Contents

1. How to use this book
2. Chapter 1. How generative AI works and its limits

   Distinguishing models, products and tools · Tokens: the units into which a model splits text · Context window: the workspace on your desk · Probabilistic generation: a very powerful autocomplete · Knowledge cutoff: the last newspaper read before leaving the country · Generation and retrieval: creating answers and looking them up · Causes and types of hallucination · Confidence in tone and strength of evidence · Reasoning models and standard models · Choosing a model tier that matches task difficulty · Multimodal input · Situations where output must not be taken at face value · Why tools should handle calculations and aggregation · Models change even under the same name: versions and reproducibility

3. Chapter 2. Effective instructions and iterative improvement

   The model is a capable newcomer without context · Start with completion criteria · Six elements of specific instructions · Role and background: what a role can and cannot provide · Explain why, and say what to do instead of only what to avoid · Providing examples: showing is faster than telling · Specifying output format: fit the next destination · Providing context by attaching material · Distinguishing reference material from instructions · Breaking work into steps · Inviting questions when information is missing, and setting priorities · Iteratively improving instructions after reviewing results · Instructing reasoning models differently · Making prompts team assets: templates and test cases

4. Chapter 3. Choosing product forms and features

   Five questions for choosing a form · Chat: a single telephone consultation · Current conversation, saved instructions, memory and project instructions · Projects (knowledge spaces): a shared team drawer · Custom assistants: a dedicated employee who knows the manual · Connectors: entering only with your own access badge · Agent features: an assistant sent on an errand · Deep research: a report comparing multiple sources · Scheduled tasks: recurring work with an alarm · API: components you assemble yourself · Workflows and agents: fixed paths and open paths · Does answering from internal documents require retraining?

5. Chapter 4. Responsible use and data

   Common prohibited categories in usage policies · High-risk domains: meaningful human review and disclosure · Do not conceal that it is AI · Five dimensions of data · Consumer and business plans · Administrator controls · Deciding whether to enter sensitive information · Identifiability: removing names is not enough · Hidden instructions in documents: prompt injection basics · Checking sources and verifying facts · Copyright and attribution are different questions · Checking bias and stereotypes · Reading beyond "not used for training" · The first 30 minutes after an incident

6. Chapter 5. Adoption value and the organizational view

   Subscription and token billing: a flat-rate plan and an electricity bill · Units to check in a pricing table · Total cost rather than unit price · Selecting use cases: one step of a task, not an entire role · Classifying by impact and risk · Compare with solutions that do not use AI · Four conditions for good adoption metrics · Baselines and measurement design · Types of metrics and their traps · Pilot sequence and criteria for expanding, revising or stopping · Organizational governance and capabilities · What if nobody uses the tools after adoption? · Avoiding dependence on one supplier

7. Appendix A. Glossary
8. Appendix B. Self-check answers and explanations
9. Appendix C. Exam day guide
10. Appendix D. Reference documents