Free Resource · Cheat Sheet

Claude AI Cheat Sheet — Write Prompts That Actually Work

Most people get mediocre results from AI because they write mediocre prompts. This cheat sheet shows you the exact difference between a weak prompt and a great one — with real example responses — plus the tricks that make Claude improve its own work.

New to AI? Not sure which assistant to use? See our comparison of Copilot, ChatGPT, Claude, Gemini & Grok.

What Claude Is Great For

  • Drafting and polishing emails, proposals, and documents
  • Summarizing and analyzing long files, contracts, or reports — it handles a lot of text at once
  • Following detailed, multi-step instructions closely
  • Careful reasoning and “think it through” problems
  • Cleaning up, organizing, or rewriting your own rough notes

The Anatomy of a Great Prompt

The single biggest upgrade to your results — give Claude four things:

IngredientExample
Role”You’re a marketing copywriter for a small IT company.”
Task”Write a 150-word LinkedIn post about why backups matter.”
Context”Our audience is Houston business owners who aren’t technical.”
Format”Friendly tone, short sentences, end with a question.”

The Prompt Ladder — Same Goal, Four Very Different Results

Here’s the part nobody shows you. The task: email a prospect who went quiet after a quote. Watch the output climb as the prompt gets better.

❌ Level 1 — Unhelpful

write a follow up email

What you’d get: A generic, hollow template — “I just wanted to follow up and see if you had any questions. Let me know!” — that sounds like everyone else’s ignored email. Claude has nothing to work with, so it guesses.

✅ Level 2 — Good (adds the task + who it’s to)

Write a follow-up email to a prospect who got a quote from us
two weeks ago and hasn't replied.

What you’d get: A polite, usable email that references the quote and asks if they have questions. Fine — but it’s still vague, has no personality, and gives the prospect no reason to reply now.

✅✅ Level 3 — Better (adds role, context, and a goal)

You're a friendly sales rep at a Houston IT company. Write a short
follow-up email to a prospect who got our managed IT quote two weeks
ago and went quiet. Goal: restart the conversation without being pushy.
Keep it under 90 words, warm and human.

What you’d get: A tight, on-brand email with a real subject line, an easy out, and one specific reason to re-engage — “Totally understand if the timing shifted. If it’d help, I can send a one-page summary of what’s included so it’s easy to compare.” Genuinely sendable.

🏆 Level 4 — Amazing (specifics + format + constraints + a soft offer)

You're a friendly sales rep at a Houston IT company (Braintek).
Write a follow-up email to "Mark," owner of a 40-person accounting
firm, who received our managed IT quote two weeks ago and went quiet.

Context: his main pain was slow response times from his last IT guy.
Goal: restart the conversation, address that pain, give an easy yes.

Constraints:
- Under 90 words, 5th-grade reading level, warm but not salesy
- Subject line + body
- One specific value reminder (our 5-minute response time)
- A low-friction CTA (offer a 10-min call OR a one-page summary)
- No "just checking in" and no guilt-tripping

Then give me a shorter second version I can use as a text message.

What you’d get: A polished email and an SMS version. The email opens by acknowledging his timing, reminds him of the 5-minute response promise that fixes his exact pain, and closes with a choose-your-own CTA. The text version is two sentences he can fire off from his phone. This is the difference between AI as a toy and AI as a teammate.

The lesson: every rung you add — role, context, the real goal, constraints, format — removes a guess Claude has to make. Fewer guesses = better output.

Make Claude Improve Its Own Work

The pros don’t write one perfect prompt — they make the AI sharpen itself. Steal these:

  • Let it upgrade your prompt first: “Before you answer, rewrite my prompt to be clearer and more complete, then answer the improved version.”
  • Make it interview you: “Ask me 3 questions you need answered to do this really well, then wait for my replies.” (Kills the guessing.)
  • Force self-critique: “Now critique that draft like a tough editor and list 3 weaknesses.” then “Rewrite it fixing those.”
  • The 1–10 trick: “Rate that answer 1–10 for a skeptical business owner, explain the score, then give me the 10/10 version.”
  • Surface hidden assumptions: “What did you assume that might be wrong here?”
  • Give it an example to match: Paste one email you love and say “match this tone and structure.”
  • Chain it: “Step 1: outline. Stop and show me. Step 2: draft only after I approve the outline.”

Claude’s Power Features Worth Knowing

  • Projects (Claude.ai): keep related chats, files, and instructions together for an ongoing task — Claude remembers the context across the project.
  • Big documents: paste long contracts, transcripts, or reports and ask for a summary, the risks, or an action list. It holds a lot of text at once.
  • Artifacts: ask for a document, table, or simple web page and Claude builds it in a side panel you can edit and reuse.
  • File uploads: drop in a PDF, spreadsheet, or image and ask questions about it.

Smart Business Uses

  • Turn a messy meeting transcript into clean notes and a task list
  • Draft first-pass replies to common customer emails (then make it critique and tighten them)
  • Summarize a long vendor contract and flag what to watch for
  • Write job descriptions, SOPs, or training docs from a few bullet points
  • Repurpose one blog post into a newsletter and three social posts

The One Rule That Matters Most

Never paste confidential or regulated data — client financials, patient records, passwords, or anything covered by HIPAA, FTC Safeguards, or an NDA — into a public AI tool. Treat it like a smart but public intern. Using AI on sensitive data safely needs a proper, secured setup — which is exactly what we help businesses build.

Need help using AI and Automation in your business?

We help Houston and DFW businesses adopt AI and automation safely — without leaking sensitive data or wasting money on tools that don't fit. Tell us what you're trying to do and we'll show you the smart way to do it.

By submitting, you agree to be contacted by Braintek about your inquiry.

FAQs

Is it safe to use Claude for business?

It can be — but never paste confidential client data, passwords, or regulated information into any public AI tool. For business use, set clear rules and consider a managed AI setup. We help businesses adopt AI safely.

What is Claude best at?

Writing and editing, summarizing and analyzing long documents, working with large amounts of text at once, careful reasoning, and following detailed instructions closely.

How do I get better answers from Claude?

Give it a role, the task, context, and the format you want — then ask it to critique or improve its own answer. The prompt ladder in this cheat sheet shows the exact difference between a weak prompt and a great one.

Ready for IT that just works?

Book a no-pressure discovery call. We'll review your setup and show you exactly where you stand.