This blog post is Human-Centered Content: Written by humans for humans.
If you’ve ever typed something into Claude and gotten back a response that was technically “correct” but not at all what you meant, welcome to the club. It’s one of the most common questions I get from folks who are new to using AI in their work.
After spending time working with Claude, I’ve learned that the quality of the output is largely dependent on the quality of what you put in. Let’s break down what AI prompting actually is, what it isn’t and how to build a prompt that gets you where you want to go.
What Prompting Is:
A skill and a tool — Getting great results from Claude is a skill that takes practice to master. Think of it like working with a very capable (but very new) colleague. They’re smart, but they need context to do their best work.
An Iterative Process — Your first prompt is rarely the end of the conversation. Working with Claude is a back-and-forth process: You get a response, you refine your ask, you try again. The best prompts I’ve written usually came after at least one round of “not quite, but closer.”
An assignment of roles and constraints — Claude responds really well to being given a clear role and some guardrails. Something like “act as a senior copywriter” or “respond in under 100 words” can completely transform your output. And don’t be shy about being specific. You aren’t bossing the AI around, you’re helping it help you.
What Prompting Isn’t:
Magic — There’s no secret phrase that unlocks perfect results every time, and anyone who tells you differently is selling something. What works for one project won’t necessarily work for another.
A search query — This is probably the most common misconception I see. Typing “quarterly report summary” into Claude is not a prompt that’s going to produce anything useful. Claude needs context, structure, and direction to give you something genuinely useful.
Mind-reading — Claude cannot infer what you meant if you didn’t say it. It doesn’t know your audience, your tone preferences, your company’s jargon or that you always want a bullet point summary at the end. If it matters to you, it needs to be in the prompt.
The Anatomy of a Great Prompt
There are a lot of guides and mnemonic hooks out there to help you create a great prompt. The CRAFT model is the basic framework I use.
- Context — What role should Claude take on, and what’s the situation? (“You are a marketing writer helping me draft an email for a product launch…”)
- Rules — What should it avoid? Are there constraints to follow? (“Do not use jargon. Keep the response under 200 words. Do not make specific pricing claims.”)
- Action — What exactly do you want Claude to do? (“Draft an email to potential customers”)
- Format — What should the output look like? A table, a bullet list, a full paragraph, a numbered checklist? (“Write an email subject and three-to-five sentence email body…”)
- Target — Who is the audience? (“…for a non-technical audience who has never heard of our product.”)
Put all these pieces together and your prompt goes from “write me a marketing email” to something Claude can run with:
You are a marketing writer helping me draft an email for a product launch. Draft an email to potential customers with a subject and three-to-five sentence email body for a non-technical audience who has never heard of our product. Do not use jargon and keep the response under 200 words. Do not make any specific pricing claims.
AI prompting is a learnable skill that is worth your time to develop. If you’re curious about how to put Claude to work for your team or want help getting started, reach out!
