Your First Effective Prompts — AI~Foundations

Why This Matters

Most people prompt AI the way they type a Google search — short, keyword-heavy, vague. This produces mediocre results and leads people to conclude that "AI isn't that useful." The problem isn't the AI. It's the input. Prompting is a skill, and like all skills, it has a learning curve. This lesson gives you the fundamentals: three changes to how you write prompts that will immediately produce better results.

The Concept

Core Reading

Tier 1 — Orientation 13 min

Remember the mental model from Lesson 1: you're not asking a question of a knowledgeable entity. You're providing context to a prediction engine. The quality of what you get back is a direct function of the quality and specificity of the context you provide.

When you type "write me a summary," the AI makes dozens of decisions on your behalf: How long? For what audience? Of what? In what tone? It will guess — often plausibly, often wrong for your specific needs. Give it those decisions upfront and you get something much closer to what you want on the first attempt.

The Three Fundamentals

1

Role — tell AI who it's being

Role specification implicitly imports a vast set of contextual assumptions — vocabulary, tone, framing, depth — that would take many sentences to specify individually.

❌ Weak"Explain machine learning to me."

✅ Strong"You are a patient teacher explaining machine learning to a curious professional with no technical background. Use everyday analogies and avoid jargon."

2

Context — give it what it needs

AI cannot read your mind, your files, or your organizational context. The most common reason AI gives unhelpful responses is missing context. Before sending, ask: what does it need to know that it might not have?

❌ Weak"Rewrite this email to be more professional."

✅ Strong"Rewrite this email to be more professional. The recipient is a new client we haven't met yet. I want to sound warm but authoritative. The current version feels too casual. Keep it under 150 words. [paste email]"

3

Format — specify what you want back

By default, AI responds in the format that seems most typical for your request type. You can and should specify what you actually need:

"Give me three options" — prevents AI from committing to one approach
"Use bullet points, not paragraphs" — for scannable reference material
"Keep the total response under 200 words" — for tight constraints
"Include a brief explanation of your reasoning" — for understanding, not just output

The Before/After Habit

Before sending a prompt, run through this quick checklist:

Have I given it a role or at least a framing for who it should be?
Have I provided all the context it needs to actually help me?
Have I specified what I want the output to look like?

You don't need all three every time. But running through it before sending will improve your first-attempt results significantly.

🔁

Iteration is not failure. Expert AI users iterate — they send a first prompt, evaluate the output, and follow up: "Make it shorter," "Less formal," "Expand the third option." Multi-turn conversations are often more productive than trying to write the perfect single prompt. Think of it as a dialogue, not a command.

Before and after: the same request, rewritten

The same request written at three levels of specificity. Notice how each version gives the prediction engine more to work with.

1

Typical first attempt

"Help me write a LinkedIn post."

The AI has no idea what topic, tone, audience, or goal you have. It guesses all of these. The result will be usable at best, forgettable always.

2

With role and context

"I'm a marketing manager who just completed a three-month AI implementation project at my company. Help me write a LinkedIn post about what we learned."

Better. But still no format spec, no audience, and "what we learned" is vague. Reasonable output — probably too long and too generic.

3

With role, context, and format

"I'm a marketing manager who just completed a three-month AI implementation project. The biggest lesson: the hardest part wasn't the technology — it was getting the team to change their workflows. Write a LinkedIn post that shares this insight for other managers going through the same thing. First-person narrative tone, start with a specific moment rather than a general statement, under 250 words."

Now the AI has enough to produce something genuinely useful on the first attempt.

Hands-On Exercise

Rewrite your weakest prompt

ClaudeChatGPTGemini

Think of a time you used AI and the result was disappointing — too generic, too long, wrong tone, missed the point. If nothing comes to mind, use one of the "weak" prompts from this lesson.

Take that prompt and rewrite it three times, adding one element each time:

Rewrite 1 — Add a role

Start with: "You are a [role]..."

Rewrite 2 — Add context

Add: audience, constraints, background, what you've already tried.

Rewrite 3 — Add a format spec

Add: length, structure, tone, output style.

Try all three versions in an actual AI tool. Which changed the result most — the role, the context, or the format spec?

Don't try to write the perfect prompt on the first try. The goal is to notice what each element adds.

Active Recall

Before moving on — close this lesson and answer these from memory. Then come back and check. Testing yourself (not re-reading) is how this sticks.

1 What are the three fundamental elements of a well-structured prompt? Give a brief description of what each one does.

2 Write a before-and-after example of a prompt from your own work or life. What specific changes did you make, and why?

3 Why is iteration a normal and expected part of good AI use, not a sign that something went wrong?

Reflection

💭

Look back at your last three AI interactions. Without judging yourself, notice: did you provide a role? Did you give context? Did you specify the output format? What one habit, if you added it to every prompt, would most improve your results?

Key Takeaway

Effective prompts give AI a role, provide necessary context, and specify the desired output format. You are not asking a question — you are providing context to a prediction engine. Iteration is normal. The goal is better first attempts, not perfect first attempts.