Summary  
This chapter introduces a four-part prompt structure—Context, Instruction, Format, Constraint—to precisely guide a generative model’s output and shows how tweaking each part affects tone, structure, and content.  

General domain of usage  
Summarizing meeting transcripts into actionable summaries for team follow-up.

You know the recipe: **Context · Instruction · Format · Constraint**. 

Now apply it to a real task — turning a messy meeting transcript into something your team can act on.


## Your Input

>[Tuesday, 10:03 — Q4 Marketing Sync] 
Sarah: Launch landing page is still pending Mike's review. Mike, when? 
Mike: Thursday morning. Sarah, drop the staging link in my DMs by Wed EOD. 
Sarah: Will do. 
Sarah: Email sequence — we agreed three, but Priya pushed back, said two. 
Priya: Right. The third was the case-study email; Acme's quote isn't in yet. 
Sarah: So two for launch? 
Priya: Yes. Third goes out two weeks later, once Acme's quote lands. 
Mike: What's the paid budget for week one? I keep seeing different numbers. 
Sarah: $8K, confirmed. 
Mike: Static creatives or the new motion ones? 
Priya: Static. Motion isn't tested yet on LinkedIn. 
Sarah: Heard from legal on the compliance disclaimer? 
Mike: No. I'll ping Tom this afternoon. 
Sarah: Cool. See you Thursday.

## Task

Write a prompt that uses all four parts. Use this as your starting point:

>I'm a marketing PM. The transcript below is from our Q4 launch sync. The audience is teammates who skipped the call. 
> 
>Summarise it so anyone can act on the outcomes. 
> 
>Reply in three bulleted sections: Decisions Made — Action Items (each starts with the owner's name in bold, then the action, then the deadline if one was given) — Open Questions. 
> 
>Don't include filler chitchat. Don't invent owners or deadlines. Mark anything unclear as TBD. 
> 
>[paste transcript]

Send the prompt, read the output carefully, then try at least one variation below.

## Experiment — This Is Where It Clicks

Change just **one part** and observe what shifts.

**1. Drop the constraint**

Remove "Don't invent owners or deadlines" — does the model start guessing?

**2. Change the format**

Ask for a table instead:

>Type | Item | Owner | Deadline

Same information, different structure.

**3. Change the context**

Set the audience to CEO instead of team — notice how the tone becomes more concise and high-level.

## Key Insight

Each part of the recipe controls something specific:

- **Context** — tone and relevance;
- **Instruction** — task clarity;
- **Format** — structure of output;
- **Constraint** — prevents mistakes.

You don't need a better model. You need a better prompt structure.

A hands-on, beginner-friendly course designed for non-technical business users to understand and leverage AI tools and concepts in their daily work. The course covers foundational AI concepts, effective communication with AI, practical use of AI-powered tools, and building simple AI assistants.

Introduction to artificial intelligence, large language models, and essential AI vocabulary for business users.

Learn how to communicate effectively with AI systems through prompt engineering and tone selection.

Explore how AI can be integrated with business tools for research, note-taking, and data analysis.

Learn the basics of agent personas and how to build and tune simple AI assistants for work.