How to Write Your First AI Prompt

Why Prompting Is the Most Valuable AI Skill You Can Learn

Here is a stat that should get your attention: 92% of Fortune 500 companies are now using AI tools in their day-to-day operations, yet internal surveys consistently show that fewer than 15% of employees feel confident writing prompts that produce genuinely useful results. That gap is where the opportunity lives.

The quality of your AI output depends almost entirely on how you communicate with the model. A vague prompt produces a vague answer. A precise, well-structured prompt produces output that saves you hours of work. This tutorial will teach you a repeatable framework — the CRISP method — that works across every AI tool and every profession.

By the end of this guide you will be able to write prompts that consistently deliver useful, specific, ready-to-use results on the first or second try.

The Before & After: Real Prompts Across Three Professions

Before diving into the framework, let's see what "good prompting" actually looks like in practice.

Doctor — Patient Discharge Summary

Before (vague prompt):

Write a discharge summary for a patient.

After (CRISP prompt):

You are a senior internal medicine physician at an academic medical center.

Write a discharge summary for a 62-year-old male patient admitted for
community-acquired pneumonia. Hospital stay: 5 days. Treatment: IV
ceftriaxone transitioned to oral amoxicillin-clavulanate. Key labs at
discharge: WBC 7.2, CRP 12 (down from 89). O2 sats stable on room air
for 48 hours.

Format: standard hospital discharge summary with sections for
Admitting Diagnosis, Hospital Course, Discharge Medications, Follow-Up
Instructions, and Red-Flag Symptoms for the patient.

Keep clinical language in the Hospital Course section but use plain
language (8th grade reading level) in the patient-facing sections.

The vague prompt produces a generic template. The CRISP prompt produces a draft that is 90% ready for chart review.

Lawyer — Contract Clause Analysis

Before:

Review this contract clause.

After:

You are a commercial contracts attorney with 12 years of experience
in SaaS agreements.

Analyze the following limitation-of-liability clause. Identify:
1. Any terms that are unusually favorable to the vendor
2. Missing carve-outs that are standard in enterprise SaaS contracts
3. Specific language changes I should negotiate

Present your analysis in a table with columns: Issue, Risk Level
(High / Medium / Low), Suggested Revision.

Clause:
"""[paste clause here]"""

Marketer — Campaign Brief

Before:

Write a marketing campaign plan.

After:

You are a growth marketing strategist at a B2B SaaS company.

Create a 4-week product launch campaign brief for our new AI-powered
analytics dashboard. Target audience: VP-level decision makers at
companies with 200-2000 employees. Budget: $15K. Channels: LinkedIn
ads, email nurture sequence, one webinar.

For each week, provide: theme, key message, channel mix, primary CTA,
and success metric. Format as a markdown table.

Tone: authoritative but not stuffy. Avoid buzzwords like "synergy"
and "leverage."

Notice the pattern: every effective prompt includes who the AI is, what you need, how to format it, and specific constraints. That is the CRISP framework.

Step 1: Define Your Goal Before You Type

The single biggest mistake beginners make is opening their AI tool and starting to type before they know what they want. Spend 30 seconds answering one question:

"If this prompt works perfectly, what specific output am I holding in my hands?"

Be concrete:

• Not "help with my presentation" but "a 5-slide outline for a board meeting about Q3 revenue growth"
• Not "write an email" but "a 150-word follow-up email to a prospect who attended our webinar but didn't book a demo"
• Not "explain this code" but "a plain-language summary of what this Python function does, including edge cases"

Quick exercise: Before your next prompt, write down the desired output in one sentence. This alone will improve your results by 50%.

Step 2: Set the Context (the C and R in CRISP)

Context means giving the AI the background information it needs to produce relevant output. Role means telling it who to be.

Context Examples

I am a product manager at a Series B fintech startup.
We sell expense management software to mid-market companies.
Our main competitor is Brex. We are positioning on ease of use.

Role Examples

You are a senior financial analyst who specializes in SaaS metrics.

You are an experienced copywriter who writes for healthcare brands.
Your style is warm, clear, and jargon-free.

Why this works: LLMs have been trained on text from millions of perspectives. When you assign a role, you activate the patterns associated with that expertise. A prompt addressed to a "senior tax attorney" will produce very different output than the same prompt addressed to a "first-year law student."

Step 3: Write Clear Instructions (the I in CRISP)

This is the core of your prompt — what you actually want the AI to do. Three rules:

1. Use imperative verbs. "Write," "Analyze," "Compare," "List," "Summarize." Not "Can you maybe help me think about..."
2. One task per prompt. If you need a blog post and social media copy and an email, run three prompts. Multi-task prompts produce mediocre results across the board.
3. State what you DON'T want. Constraints are just as useful as instructions. "Do not include pricing information." "Avoid bullet points — use flowing paragraphs instead."

Examples of Strong Instructions

Summarize this 20-page research paper into 5 key findings.
Each finding should be 2-3 sentences. Include the page number
where each finding appears.

Write 3 subject line variations for a cold outreach email.
Target audience: HR directors at companies with 500+ employees.
Goal: get them to open the email. Do not use emojis or ALL CAPS.

Step 4: Add Specifics and Format (the S in CRISP)

Specifics are the constraints that turn a decent prompt into a great one:

The more specifics you provide, the less you need to iterate. Think of specifics as bumper rails in bowling — they keep the AI from going into the gutter.

Step 5: Ask for Proof or Reasoning (the P in CRISP)

This step separates amateur prompts from professional ones. When you ask the AI to show its work, the quality of the output improves dramatically:

Explain your reasoning before giving the final recommendation.

For each suggestion, include one real-world example of a company
that has done this successfully.

After your analysis, rate your confidence level (High / Medium / Low)
and explain what additional information would increase your confidence.

This is especially important for high-stakes decisions — legal analysis, medical recommendations, financial projections, strategic plans. The reasoning often reveals when the AI is uncertain or working from incomplete information.

Step 6: Iterate Like a Pro

Your first prompt rarely produces the perfect result, and that is completely normal. Professional prompt engineers expect to iterate 2-3 times. Here is how:

Iteration pattern 1: Sharpen the output

This is good, but make these changes:
- Section 3 is too technical. Rewrite it for a non-technical audience.
- Add specific dollar amounts in the ROI section.
- The conclusion should end with a clear call to action.

Iteration pattern 2: Change direction

Let me try a different approach. Instead of a formal report,
rewrite this as a 5-minute presentation script with a conversational tone.

Iteration pattern 3: Expand or compress

This is too long. Cut it to 50% of the current length while
keeping all the data points. Remove the background section entirely.

Pro tip: save your best prompts in a personal "prompt library" — a simple notes document organized by category. Within a week you will have a collection of reusable templates that make you dramatically faster.

Common Beginner Mistakes (and How to Fix Them)

Practice Exercises

Put the CRISP framework to work right now. Try these three exercises in order of increasing difficulty.

Exercise 1: Basic (5 minutes)

Rewrite this vague prompt using the CRISP framework:

"Help me write a LinkedIn post about AI."

Your rewritten prompt should include a role, specific topic, target audience, word count, and tone.

Exercise 2: Intermediate (10 minutes)

You are preparing for a job interview. Write a CRISP prompt that asks the AI to:

• Generate 5 likely interview questions for your specific role and industry
• Provide a strong sample answer for each
• Flag any common mistakes candidates make

Exercise 3: Advanced (15 minutes)

Take a real work task you need to do this week. Apply the full CRISP framework:

1. Write the prompt
2. Run it through your AI tool
3. Iterate at least twice based on the output
4. Compare your final output to what you would have produced without AI

Note how long the task took with AI vs. your estimate without it.

What to Learn Next

• What Is Prompt Engineering? — understand the theory behind the techniques in this tutorial
• Prompt Engineering Cheatsheet — a printable quick-reference card with all the patterns from this guide
• Automate Email Responses with AI — apply your new prompting skills to a real workflow