The CRTSE framework for prompt engineering that consistently produces professional-grade AI output. Detailed guide with before/after examples for every step.
After testing 10,000+ prompts across every major AI model, I found that output quality correlates with five specific elements. Miss any one, and quality drops from a 9/10 to a 5/10. Include all five, and you get consistently professional output.
I call it the CRTSE Framework: Context, Role, Task, Specifics, Examples.
The CRTSE Framework
C — Context
What it is: Background information the AI needs to understand your situation.
Without context:
"Write a product description for our new feature."
// Output: Generic, could be about anything
With context:
"Our company sells project management software to remote teams
of 10-50 people. We just launched a new async video messaging
feature that lets team members record and share short video
updates instead of scheduling meetings."
// Output: Targeted, relevant, specific
Rule: Include who you are, who your audience is, and what situation you’re in. More context almost always improves output.
R — Role
What it is: The expertise lens through which the AI should respond.
Without role:
"Tell me about our pricing strategy."
// Output: Generic business advice
With role:
"You are a SaaS pricing consultant who has worked with 50+
B2B companies. You specialize in value-based pricing for
companies in the $1M-$10M ARR range."
// Output: Expert-level, specific, actionable
Rule: The more specific the role, the better. “Marketing expert” is weak. “B2B SaaS content marketing strategist who focuses on developer tools” is strong.
T — Task
What it is: The specific action you want performed.
Weak task:
"Help me with my landing page."
// Too vague — help how?
Strong task:
"Write the hero section copy for our landing page. This includes:
1. A headline (under 10 words) that communicates the core benefit
2. A subheadline (under 25 words) that expands on the headline
3. A CTA button text (3-5 words)
4. 3 bullet points highlighting key features"
// Clear deliverable, specific format
Rule: Define the deliverable, not just the topic. Say what you want to receive, not what you want to discuss.
S — Specifics
What it is: Constraints, requirements, and preferences.
"Constraints:
- Tone: Professional but approachable (not corporate-speak)
- Length: 150-200 words total
- Must include the phrase 'async video messaging'
- Should address the pain point of meeting fatigue
- Don't mention competitors by name
- Use second person ('you') not third person"
Rule: Every constraint you add reduces randomness and increases relevance. Constraints are not limitations — they’re precision tools.
E — Examples
What it is: Showing the AI what good output looks like.
"Here's an example of hero copy we loved from a previous page:
'Ship faster. Sleep better. Deploy with confidence.'
'One-click deployments, automatic rollbacks, and real-time
monitoring so your team can move fast without breaking things.'
Write something with similar energy and structure for our
async video messaging feature."
Rule: One good example is worth 100 words of instruction. Show, don’t tell.
Before/After Examples
Example 1: Email Marketing
Before (no framework):
"Write a marketing email for our product launch."
After (CRTSE):
[Context] We're launching a new AI-powered feature in our
project management tool that auto-generates meeting summaries
from Zoom calls. Our users are tech-savvy project managers
at companies with 20-200 employees.
[Role] You are an email copywriter specializing in SaaS
product launch emails with high open and click rates.
[Task] Write a product launch email that:
1. Subject line (under 50 chars, creates curiosity)
2. Preview text (under 90 chars)
3. Body: hook → problem → solution → CTA
4. Total body length: 150-200 words
[Specifics] Tone: excited but not hype-y. Focus on time saved.
Include a specific stat: "saves 4.2 hours per week per PM."
CTA: "Try it free for 14 days"
[Example] Our best-performing email last quarter:
Subject: "Your Mondays just got 2 hours shorter"
[paste example body]
Comments · 0
Beta: comments are stored locally on your device and not visible to other readers.
No comments yet. Be the first to share your thoughts.