Lesson 3.2: Writing Your First Prompts

Purpose

You've chosen a tool. Now you need to know how to communicate with it. This lesson teaches "prompt construction"—not as a deep engineering discipline, but as practical communication.

By the end, you'll write clear prompts that get useful responses.

Why This Matters for Managers

Poor prompts = poor outputs = wasted time and frustration.

Bad prompt: "Write an email"

Result: Generic, forgettable email

Good prompt: "Draft an email to my team announcing a new policy. Tone should be direct but warm. Include: what's changing, why, how it affects them, when it starts."

Result: Specific, useful email you can actually use

The difference is in the detail and structure you provide. Writing good prompts is a practical skill, not magic.

Core Concepts

The Prompt Framework

A good prompt has these elements:

Element 1: The Task (What You Want)

Be specific about what you want the AI to do.

Vague: "Tell me about communication."

Specific: "Draft an email to my team about our new meeting-free Fridays policy."

Why it matters: Specificity helps AI understand exactly what you want, not something generic.

Element 2: Context (What AI Needs to Know)

Provide context the AI wouldn't know.

Without context: "Draft an email about the project status."

(AI will produce generic status email)

With context: "Our team has been working on the website redesign for 6 weeks. We're on schedule but slightly under budget. The team worked really hard. I want to thank them and keep momentum high for the final phase."

(AI produces an email that reflects your situation and goals)

Why it matters: AI doesn't know your organization, situation, team, or what you care about. You have to tell it.

Element 3: Format or Structure (How You Want It)

Specify the format or structure of the output.

Without format: "Summarize the meeting."

(AI might write a paragraph)

With format: "Summarize the meeting in 5 bullet points: (1) key decisions made, (2) action items with owners, (3) timeline for next steps, (4) risks to watch, (5) follow-up questions."

(AI produces exactly what you need)

Why it matters: Format constraints help AI understand what you're looking for and make output easier to use.

Element 4: Constraints or Preferences (What Matters to You)

Specify tone, length, audience, or other preferences.

Without constraints: "Write an update."

(Could be formal, could be casual, could be 100 words or 1000)

With constraints: "Write a 150-word update for our leadership team on project progress. Tone should be professional but not stuffy. Focus on what's been accomplished and what's coming next. Skip technical details—assume non-technical audience."

(AI writes to your constraints)

Why it matters: Constraints help AI understand your goals and avoid output you'll have to revise.

The Prompt Template

A simple template for good prompts:

[Task] + [Context] + [Format] + [Constraints] = Good Prompt

Example:

"Draft a one-paragraph email to my team thanking them for their work on the Q3 launch. [Task: draft email] We shipped on time despite the complexity, and the team handled uncertainty really well. [Context: situation they should know] Keep it warm and personal, not corporate. [Constraints: tone] Maybe 100-150 words. [Constraints: length] The goal is to make them feel appreciated and excited for the next project."

[Constraints: goal]

Iteration: The Secret to Better Prompts

Your first prompt might not be perfect. That's okay. Iteration is where prompts get better.

The iteration cycle:

1. Write a prompt
2. Get AI response
3. Evaluate: "Is this what I wanted?"
4. If no: Refine the prompt with more detail
5. Repeat

Example iteration:

First attempt:

"Draft talking points for a conversation with Sarah about her performance."

AI output: Generic performance feedback talking points.

Your evaluation: "This doesn't reflect what I actually want to say. I want to be coaching, not critical. And it misses the context that Sarah is new."

Refined prompt:

"Draft talking points for a coaching conversation with Sarah about her performance. Context: She's been in the role for 6 weeks. She's learning quickly but sometimes over-commits. My goal is to help her learn how to prioritize and say no. Tone should be supportive and developmental, not critical. Include: strengths she's showing, the specific challenge (over-committing), and questions I should ask her to help her think through solutions."

AI output: Much more targeted. Reflects your coaching approach and context.

What Makes Prompts Better (Or Worse)

What Makes a Prompt Better

1. Specific, not vague. "Draft an email about Q3 results to my team, emphasizing what we did well and what we're working on" is better than "Write something about Q3."

1. Includes context. Telling the AI your situation, your role, your team, or what you care about makes output much better.

1. Specifies format. Bullet points? Paragraph? 100 words? Tell it.

1. Shows tone. Formal? Casual? Direct? Warm? This matters.

1. Clear about the goal. Why are you communicating? What do you want the outcome to be?

What Makes a Prompt Worse

1. Too vague. "Write an email" could mean anything.

1. No context. AI produces generic output when it doesn't know your situation.

1. Contradictory. "Make it long but concise" confuses AI.

1. Asks AI to decide something only you can decide. "Should I give feedback?" (AI can't decide; you should ask it to help you think through it.)

1. Expects AI to know things it doesn't. "Write an email about our policy change" without explaining what policy changed or how it affects people.

Practical Managerial Use Cases

Case Study 1: Drafting an Email - Building a Better Prompt

Initial prompt:

"Draft an email to my team about flexible work."

Why it's weak:

• No context about what flexible work means
• No sense of what you want to accomplish
• No tone guidance
• No format
• Result: Generic corporate email

Improved prompt:

"Draft an email to my 12-person engineering team about our new flexible work policy. Context: We're moving from 'work from home 2 days' to 'mostly remote with optional in-office days.' The team wanted this. I want them to feel trusted and excited. Tone: friendly and positive, not corporate-y. Length: 1-2 short paragraphs. Include: what's changing, why we're doing it, that we trust them to do great work remotely, how to schedule office time if they want. Include a line about maintaining our team connection."

Why it's better:

• Specific about what the change is
• Clear about context (team wanted it)
• Tone guidance (friendly, not corporate)
• Format (short)
• Goal (feel trusted and excited)
• Result: Actually usable email

Case Study 2: Summarizing Information - Building a Better Prompt

Initial prompt:

"Summarize this meeting transcript."

Why it's weak:

• No guidance on what matters
• No format
• AI might summarize everything equally, missing what's important
• Result: Long, generic summary

Improved prompt:

"Summarize this meeting transcript focusing on: (1) What decisions were made and by whom? (2) What are the action items and owners? (3) What are key risks we should watch? Use bullet points. Keep it to 10-15 bullets total. Skip the small talk and background context."

Why it's better:

• Specifies what matters to you
• Format (bullets)
• Length constraint
• What to skip
• Result: Focused summary you can actually use

Case Study 3: Brainstorming - Building a Better Prompt

Initial prompt:

"What are ways to improve communication?"

Why it's weak:

• Too broad and vague
• AI might generate obvious or generic ideas
• No context about your situation
• Result: Unhelpful list

Improved prompt:

"I'm a manager of 15 engineers. Right now we have: daily standups (15 min), weekly team meetings (1 hour), 1-on-1s (monthly 30 min). The problem is engineers say they feel disconnected from strategy and don't know how decisions are made. I want ideas that don't add more meetings. What are 10 creative ways we could improve how engineers stay connected to strategy and decision-making without adding time?"

Why it's better:

• Context about your team and situation
• Clear about what's not working and why
• Constraints (no more meetings)
• Specific goal
• Result: Targeted ideas you can actually consider

Anti-Patterns / Misuse Risks

Misuse Risk 1: Over-Detailed Prompts

"I'll provide every detail so AI understands perfectly."

Why it fails: Prompts that are too long become confusing. AI gets lost in the detail.

Misuse Risk 2: Asking AI to Decide

"Should I give Sarah the promotion?"

Why it fails: AI can't decide. It can only present options.

Misuse Risk 3: Expecting AI to Know Your Organization

"Draft our Q4 strategy."

Why it fails: AI doesn't know your strategy, market, or goals.

Misuse Risk 4: Not Iterating

"The AI's output wasn't great, so AI isn't useful for this."

Why it fails: Often your prompt was unclear. Iterating usually helps.

Human Judgment Checkpoints

As you write prompts, ask:

1. Is this clear? Could another person understand what I'm asking for?
2. Did I provide context? Does AI know my situation?
3. Is this too vague? Could the AI interpret this multiple ways?
4. Am I asking AI to decide? Or to help me think?
5. Have I tried it and iterated? Or am I judging before trying?

Responsible AI Considerations

Clear Instructions

Writing clear prompts is actually an important responsibility. It helps you get better output, but it also helps you think through what you actually want.

Verification Before Using

Even with a good prompt, always verify the output. Good prompts don't guarantee accuracy.

Iterating Appropriately

Iterating is good. Expecting AI to read your mind is unrealistic. Good prompts require some back-and-forth.

Practice / Reflection Prompts

1. Your First Prompt: Write a prompt for a task you want to delegate to AI. Use the framework: task + context + format + constraints.

1. Clarity Check: Could someone else understand what you're asking for? If not, refine it.

1. Context Exercise: Pick a vague prompt. Add context that would help AI understand your situation.

1. Iteration Practice: Write a prompt, get AI response, refine the prompt, try again. Compare the two outputs.

1. Constraint Exercise: Pick a task and write two versions of the prompt: one without constraints, one with tone/length/format guidance. See the difference.

1. Decision vs. Analysis: Look at your prompts. Are any asking AI to decide something only you can decide? How would you reframe them?

Key Takeaways

1. Be specific, not vague. "Draft an email about flexible work to your team, explaining why it matters, with a warm tone" beats "Draft an email."

1. Provide context. Tell AI your situation, your team, what you care about.

1. Specify format. Bullets? Paragraphs? 100 words? It matters.

1. Indicate tone and constraints. Formal? Casual? Brief? This helps a lot.

1. Iterate, don't abandon. First response not great? Refine the prompt and try again.

1. Ask AI to help you think, not decide. "Help me think through X" is better than "Should I do X?"