AI Power User
Go from good prompts to real leverage: APIs, grounding, tools, automation and the rules that keep it safe.
This course takes a confident prompter under the hood. You will learn how models really work, write prompts that hold up under pressure, make your first API calls, ground answers in your own documents, plug tools into assistants with MCP, automate a real workflow with a review gate, and set governance rules your team can actually live with.
What you will be able to do
- Explain tokens, context windows, sampling and reasoning effort well enough to predict cost and behaviour
- Use meta-prompting, chaining and rubric-based self-critique to get reliable output on hard tasks
- Call Anthropic, OpenAI and Groq models from Python and curl, and pick the right model for a job
- Ground an assistant in your own documents with and without code, and judge whether retrieval is working
- Connect an assistant to business systems through tools and MCP, and audit what it can reach
- Build an automated workflow with an AI step, a review gate and logs, inside a POPIA-aware usage policy
Course outline
How models really work
4 lessons · quizTokens, context windows, sampling, prompt roles and reasoning effort: the four dials that explain most of what you see.
- Context windows, tokens and cost 10 min
- Temperature, sampling and why answers vary 9 min
- System prompts vs user prompts 9 min
- Reasoning models: when thinking time is worth it 10 min
- Module quiz 6 questions
Advanced prompting
5 lessons · quizTechniques that hold up on hard tasks: let the model write the prompt, break work into steps, grade with rubrics, and understand the attack that targets all of it.
- Meta-prompting: ask the model to write the prompt 9 min
- Decomposition and chaining 10 min
- Self-critique and rubric-based review 9 min
- Prompt injection: the attack you must understand 10 min
- Exercise: build and stress-test a rubric exercise · 25 min
- Module quiz 6 questions
APIs and playgrounds
5 lessons · quizMove from the chat window to the API: get keys safely, make your first calls to Anthropic, OpenAI and Groq, and choose models on evidence.
- What an API is and why you would want one 8 min
- Getting keys: OpenAI, Anthropic, Groq 9 min
- Your first API call in Python and curl 12 min
- Choosing a model: quality, speed, price 9 min
- Exercise: the same task on three models exercise · 30 min
- Module quiz 6 questions
Grounding answers in your data (RAG)
4 lessons · quizWhy models guess about your business, how retrieval fixes it, and how to build and measure it with and without code.
- Why models do not know your data 8 min
- Embeddings and vector search in plain terms 9 min
- RAG without code: Projects, NotebookLM and file search 9 min
- RAG with code: a minimal pipeline and how to evaluate retrieval 14 min
- Module quiz 6 questions
Tools, function calling and MCP
4 lessons · quizHow a model asks your code to act, how MCP standardises the plug, and how to connect real business systems without handing over the keys.
- Tool use: letting the model act 10 min
- Model Context Protocol: a standard way to plug tools into assistants 10 min
- Connecting assistants to business systems safely 10 min
- Exercise: connect one MCP server and audit what it can do exercise · 30 min
- Module quiz 6 questions
Automation
4 lessons · quizPut an AI step inside a real workflow: pick a platform, design the step as a contract, handle failure, keep a human at the gate and log everything.
- No-code automations: Zapier, Make, n8n and Power Automate 9 min
- Designing an AI step inside a workflow 10 min
- Error handling, human review and logging 10 min
- Exercise: one automated workflow with a review gate exercise · 40 min
- Module quiz 5 questions
Governance and responsible use at work
3 lessons · quizA one-page policy people follow, data classification under POPIA, and what to ask vendors about plans, residency and audit.
- An AI usage policy people actually follow 9 min
- Data classification and POPIA in practice 10 min
- Enterprise plans, data residency and audit 10 min
- Module quiz 6 questions