Most people use AI like a search engine. Here's how to use it like a $200/hr expert.
You've been using ChatGPT or Claude for months. But your results still feel... generic. You spend more time rewriting AI output than you save by using it.
The problem isn't the AI. It's the prompt. And fixing it takes less than 30 seconds — once you know the techniques.
Here are 5 prompt engineering techniques that separate power users from everyone else.
1. Set a Role (The AI "Mode Switch")
When you tell an AI "You are a senior copywriter with 15 years of experience in B2B SaaS," something shifts. The vocabulary changes. The recommendations get more specific. The output matches what that professional would actually produce.
Why it works: AI models learned from millions of examples of how different professionals write and think. Setting a role activates that pattern.
Bad: "Write me an email."
Good: "You are a senior account manager at a marketing agency. Write a client check-in email that addresses a missed deadline diplomatically."
2. Front-Load Context
The single biggest lever for prompt quality. Most people assume the AI will "figure it out." It won't. Everything you know about your business, audience, and goals — the AI knows none of it.
Give it:
- Who your audience is
- What your business does
- What you've tried before
- Relevant constraints (budget, timeline, tone)
Rule of thumb: You almost can't give too much context. The AI will use what's relevant and ignore the rest.
3. Specify the Format
AI defaults to walls of text. If you want bullet points, a table, specific headers, or a certain length — say so.
Format the output as a table with columns:
Feature | Benefit | One-liner for the website
Include 10 rows. Keep each one-liner under 15 words.
This one change saves you 80% of your editing time.
4. Use Constraints to Kill Generic Output
Constraints are the secret weapon. When you tell the AI what NOT to do, it gets creative.
- "Do NOT use the words: easy, simple, powerful, streamline"
- "Every sentence must be under 15 words"
- "Write as if the reader has already failed at this 3 times"
Constraints eliminate lazy defaults and force originality.
5. Iterate, Don't One-Shot
The best output never comes from a single prompt. Treat it like directing:
- Get a draft
- "Good, but too formal. Make it more casual."
- "Keep paragraphs 1 and 3, rewrite paragraph 2."
- "Now cut it in half."
Four prompts that take 2 minutes will produce better output than one "perfect" prompt you spend 10 minutes crafting.
The Framework Behind All 5: RCTFE
These techniques combine into what I call the RCTFE framework:
- Role — Who is the AI acting as?
- Context — What background does it need?
- Task — What exactly should it produce?
- Format — What should the output look like?
- Examples — What does "good" look like?
Use all five in every prompt, and your results will be dramatically better.
Real Example Using RCTFE
Here's a prompt using all 5 elements:
ROLE: You are a senior email marketer who specializes
in welcome sequences for SaaS products.
CONTEXT: My product is TaskFlow, a project management
tool for freelancers. Users are non-technical (mostly
designers and writers). They signed up because they're
overwhelmed by complex tools. Our brand voice is
friendly, simple, encouraging.
TASK: Write the first email in a 5-email welcome
sequence. Make new users excited, show them the ONE
most important feature (daily priority view), and get
them to complete their first task.
FORMAT: Subject line (+ 2 alternatives), preview text,
body under 200 words, CTA button text, P.S. line.
EXAMPLE of our voice: "Hey! So glad you're here.
TaskFlow was built for people like you — creative pros
who'd rather be doing great work than wrestling with a
project board."
The output from this prompt is dramatically better than "Write a welcome email for my app."
Try RCTFE on your next prompt and see the difference.
I've compiled 150+ prompts using this framework for different business functions — marketing, sales, content, strategy, email, social, operations. The full PromptCraft Pro guide and toolkit is now pay-what-you-want (from $1).
Top comments (0)