Course handout
A quick reference from AI Productivity without Risk: what generative AI is, how to ask for what you need, how to check what comes back, and the ethical lines to hold. Pick a section below or from the Reference menu.
Software that creates new content — text, images, audio, video, and code — by learning patterns in massive datasets and predicting a useful response, rather than retrieving existing information.
It's probabilistic, not a search engine: fluent output isn't always accurate, and it always needs human judgment.
Source: IBM Research, “What is Generative AI?” (ibm.com/think, 2024)
A simple recipe for safe and effective prompting.
| Step | What to include | Example |
|---|---|---|
| A Ask Specific task | The clearest, most actionable ask possible | “Rewrite this email to be more persuasive for a skeptical CFO.” |
| B Background Thorough context | Who you are, who the AI is, relevant background | “I'm a marketing manager preparing for a board meeting…” |
| C Constraints Important limits | Hallucination guardrails, sources to use and avoid, full-file inputs, what not to do | “Only use the attached report. If unsure, say so. Don't exceed 1 page.” |
| D Deliverable Clear output format | Length, structure, tone, style; include an example of good output | “3 bullet points, plain language, no jargon, under 100 words.” |
These prompts are the basic vocabulary for working with AI.
| Prompt | Starter phrase | When to use |
|---|---|---|
| Understand | ||
| Explain | “Explain this in plain English…” | Learning, simplifying, decoding jargon |
| Summarize | “Summarize this in 5 bullets…” | Condensing emails, articles, meeting notes |
| Create | ||
| Draft | “Draft a professional email/document that…” | Creating first drafts of emails, letters, proposals |
| Ideate | “Give me 10 ideas for…” | Generating options, titles, names, plans |
| Plan | “Create a step-by-step plan to…” | Turning a goal into actions and milestones |
| Transform | ||
| Rewrite | “Rewrite this to be clearer, warmer, more concise…” | Improving tone, clarity, professionalism, brevity |
| Structure | “Turn this into an outline / table / checklist…” | Making messy thoughts usable and actionable |
| Assess | ||
| Compare | “Compare A and B in a table, with pros and cons…” | Evaluating options, tools, products, strategies |
| Critique | “Review this and tell me what's unclear or risky…” | Getting feedback before sending or publishing |
| Evaluate | “Score this against a rubric of…” | Structured assessment against criteria |
Go from AI response to trusted answer with these five steps.
Does this answer matter enough to verify?
What facts, numbers, names, dates, quotes, or recommendations is the AI asking me to believe?
Where did this information come from, and is that source reliable?
Does it match trusted documents, official sources, current data, or what I already know to be true?
Can I use it as-is, revise it, verify further, or discard it?
Try these constraints.
Try these constraints.
For twenty-four more targeted constraints, see the Guardrails page.
Guiding principle: Do no harm
| UNESCO core value | What it means for generative AI use |
|---|---|
| Human rights, fundamental freedoms & human dignity |
|
| Environment & ecosystem flourishing |
|
| Diversity & inclusiveness |
|
| Living in peaceful, just & interconnected societies |
|
Source: UNESCO, Recommendation on the Ethics of Artificial Intelligence, adopted 23 November 2021. “What it means for generative AI use” was created by Claude and the author.
