Prompt Engineering & Generative AI at Work

A free, practical, trilingual course on using large language models (LLMs) such as ChatGPT, Claude, Gemini and Microsoft Copilot productively and safely at work. Learn how generative AI really works, master proven prompt-engineering patterns (zero-shot, few-shot, chain-of-thought, role and structured prompts), ground answers with retrieval-augmented generation (RAG), evaluate and iterate on outputs, and apply responsible-AI guardrails for privacy, bias, hallucination and security. Includes rich visual lessons, original practice questions and a 10-question final exam (80% to pass). The course is organized into 7 modules, ending with a final exam (pass mark 80%). It is free awareness-level training designed for anyone who needs a practical, working understanding of the topic.

What you'll learn

• Module 1 — How Generative AI & LLMs Work
• Module 2 — Anatomy of a Good Prompt
• Module 3 — Core Prompting Patterns
• Module 4 — Parameters, System Prompts & Context
• Module 5 — Grounding with RAG & Custom Data
• Module 6 — Evaluating & Iterating on Outputs
• Module 7 — Responsible & Secure Use at Work

Learning objectives

• Explain in plain terms how large language models generate text: tokens, embeddings, context windows and next-token prediction
• Recognise the strengths and limits of generative AI, including hallucination, bias, knowledge cut-off and non-determinism
• Write clear, structured prompts using role, task, context, format and constraints
• Apply core prompting patterns: zero-shot, few-shot, chain-of-thought, role prompting and decomposition
• Tune model behaviour with inference parameters (temperature, top-p, max tokens) and a system prompt
• Ground answers in your own data with retrieval-augmented generation (RAG) instead of fine-tuning where possible
• Evaluate, test and iteratively improve prompts and outputs with human review and simple metrics
• Apply responsible-AI guardrails for privacy, security (prompt injection), fairness and human oversight
• Sit a 10-question final exam (80% to pass) built from original, scenario-based questions

Course modules in brief

Module 1 — How Generative AI & LLMs Work

Generative AI produces new content — text, code, images, audio — by learning statistical patterns from huge datasets. The text engines you prompt at work are large language models (LLMs) such as GPT, Claude and Gemini. They are built on the transformer architecture, whose self-attention mechanism weighs how much each word relates to every other word in your input.

Module 2 — Anatomy of a Good Prompt

A prompt is the instruction you give the model. Vague prompts get vague answers. Strong prompts are specific, structured and contextual . A reliable template combines five ingredients: a role , a clear task , the context the model needs, the desired format , and explicit constraints .

Module 3 — Core Prompting Patterns

A handful of reusable patterns cover most professional needs. Knowing when to reach for each one is what separates casual users from skilled ones.

Module 4 — Parameters, System Prompts & Context

Beyond the words you type, a few inference parameters shape every response. They control the trade-off between focused, repeatable output and creative, varied output.

Module 5 — Grounding with RAG & Custom Data

Out of the box, an LLM only knows its training data and what you paste into the prompt. To make it answer reliably from your documents — policies, manuals, a knowledge base — there is a ladder of customisation, from cheapest to most involved. Climb only as high as the task needs.

Module 6 — Evaluating & Iterating on Outputs

Prompting is empirical: you rarely get the best prompt on the first try. Treat it as a test-and-refine loop . Write a first prompt, run it, inspect what is wrong, change one thing, and run again — keeping the versions that work.

Module 7 — Responsible & Secure Use at Work

Using generative AI at work carries real risks. Knowing which ones are most likely and most damaging tells you where to put guardrails. The matrix below maps common LLM risks by likelihood and impact.

Exam and certificate

The course has 7 modules and ends with a final exam; the pass mark is 80%.

Frequently asked questions

Is this course free?

A free, practical, trilingual course on using large language models (LLMs) such as ChatGPT, Claude, Gemini and Microsoft Copilot productively and safely at work.

What does “Module 1 — How Generative AI & LLMs Work” cover?

Key vocabulary: a token is a chunk of text (roughly a word-piece) the model reads and predicts; an embedding turns tokens into numeric vectors so similar meanings sit close together; the context window is how many tokens the model can hold at once (your prompt plus its reply). Generation is autoregressive — the model predicts the next token, one at a time, sampling from a probability distribution.

What does “Module 2 — Anatomy of a Good Prompt” cover?

Two structural habits raise quality fast. First, put instructions before the data and use clear delimiters (triple quotes, XML-style tags, or headings) so the model can tell instructions from content. Second, ask for the output schema you will actually use — a table, JSON, or numbered list — which makes results easier to parse and reuse.

What does “Module 3 — Core Prompting Patterns” cover?

Few-shot prompting teaches by example: include a few demonstrations of the exact transformation you want, and the model imitates the pattern. Chain-of-thought asks the model to reason out loud, which markedly improves accuracy on arithmetic, logic and multi-step problems — though it costs more tokens and you may hide the reasoning in the final output.