How to use Claude without wasting time
Claude is Anthropic's assistant, and it's built for real work. It writes well. It'll read a long contract without losing the thread. Ask a lazy question and you'll get a lazy answer back. Give it the goal plus one worked example, like a sample email you liked.
1. Tell it who it should be
Give Claude a job. That's the fastest way to change the answer. Try "You are a careful bookkeeper" before asking it to clean up a messy invoice note. Or use "You're a blunt editor" before pasting a sales page. One short sentence changes the tone and what it notices.
2. Give it the context it can't guess
Claude doesn't know your shop. It doesn't know your customers or last week's meeting. Paste the background: what you're trying to do, who it's for, and which facts matter. A request with a product name, buyer type, and deadline beats "write me a plan" every time.
3. Ask for the exact format you want
Don't make it guess. If you don't name the shape, Claude usually gives you tidy paragraphs. Ask for five bullet points. Ask for a table. Say "Keep it under 100 words" or "Write it as an email." You'll skip a cleanup round.
4. Show it one good example
Show the target. One good example beats a pile of adjectives. Paste an email you liked and say "match this tone." Show a product description you're happy with and say "write three more like this." Claude learns faster from a real sample than from words like polished or friendly.
5. Use its real strength: long documents
Use the big window. Claude is unusually good at reading long things. Paste a contract, a report, or a big email thread. Ask for the summary, the key risks, or the three decisions buried inside it. You'll turn a wall of text into the two minutes that matter.
One caution: don't paste anything confidential — customer data, contracts, secrets — into any AI tool until you've checked its privacy settings and terms. We'll cover that properly in the free Learn AI guide. Don't wing it.
6. Work in passes, not one big shot
Don't chase perfect. Get a rough draft, then steer it. Say "shorter," "less formal," "add a section on pricing," or "cut the last paragraph." Treat Claude like a fast writer sitting next to you. Three quick passes beat one enormous prompt.
7. Make it check its own work
Make it argue back. After Claude answers, ask "What might be wrong or missing here?" Or ask "What would a skeptic say about this?" It'll often catch weak spots, bad assumptions, or claims that sound too certain. It's a good habit, but it is not a guarantee.
8. Save the prompts that work
Keep the good ones. When a request gives you a strong result, paste it into a notes doc with a title. After a few weeks, you'll have a small library of "known-good" prompts. Your team won't start from a blank page every Monday.
What Claude can't do
Know the limits. Getting good with Claude also means knowing where you can't trust it:
- It can be confidently wrong. It'll invent facts, names, numbers, and even quotes, then say them like a receipt total. Always verify anything that has to be exactly right.
- It doesn't know today's news unless it's connected to the web in that moment. Its built-in knowledge has a cut-off date, so don't ask it for fresh election results without live sources.
- It's not your lawyer, doctor, or accountant. It's useful for a first draft or a clear explanation. Don't let it make the final call on anything serious.
Use Claude this way and it stops feeling like a toy. Give it a clear role, real context, one example, the format you want, and a couple of passes. You'll save time on drafts, summaries, and cleanup work. No tricks. Just give the assistant enough to work with, then check what matters.