Most people talk to Claude the same way they’d type into a search bar: short, vague, one-shot. That works fine for quick facts. But if you’re using Claude for real work — drafting documents, thinking through a decision, writing code, analyzing data — a few small habits make the difference between an answer that’s “fine” and one that actually saves you time.
Here are five that consistently make the biggest difference.
1. Say what “good” looks like, not just what you want
“Write me a product description” gets you something generic. “Write me a product description for a $40 ceramic mug, aimed at people buying a housewarming gift, under 60 words, warm but not cutesy” gets you something you can actually use.
The gap between those two prompts isn’t length for its own sake — it’s that the second one gives Claude a target to aim at: audience, constraints, and tone. When you’re not sure what “good” looks like yourself, that’s fine too — just say so, and Claude can offer a few directions to choose from rather than guessing at one.
2. Use Projects for anything you’ll come back to
If you’re working on something over multiple sessions — a client account, a research thread, an ongoing piece of writing — a Claude Project is worth setting up. A Project bundles together custom instructions, reference files, and a set of related conversations, so you’re not re-explaining context every time you open a new chat.
The habit that actually pays off here: put the real context in the instructions and files, not just the project name. A project titled “Marketing” tells you nothing six months from now, and — more importantly — Claude doesn’t read the title at all. The substance needs to live in what you actually write into the project.
3. Let Claude show its work when the task is complex
For anything with more than one moving part — a multi-step analysis, a tricky piece of code, a decision with tradeoffs — ask Claude to reason through it step by step before landing on an answer, rather than jumping straight to a conclusion. You’ll catch mistakes faster, and you’ll often find the reasoning itself is as useful as the final answer.
This matters more than people expect for anything involving numbers, logic, or multi-part instructions, where skipping straight to “the answer” is exactly where errors hide.
4. Give examples instead of just describing style
If you want a specific tone — formal, playful, terse, whatever — showing Claude an example does more work than describing it. “Write it like this: [paste a paragraph you like]” consistently outperforms “write it in a professional but friendly tone,” because tone is easier to demonstrate than to define.
This is especially useful for recurring tasks: emails in your voice, code in your team’s style, social posts that match your brand. Once you’ve found a good example, save it in a Project’s files so you’re not re-pasting it every time.
5. Push back and iterate — don’t start over
If the first draft isn’t right, the fastest fix is usually not a brand-new prompt from scratch. Tell Claude specifically what’s off: “make the second paragraph shorter,” “this is too formal,” “you missed the part about budget.” Claude has the full conversation in view and can make a targeted edit, which is almost always faster than re-explaining the whole task.
For code, documents, and other files, this is where Artifacts help — content Claude creates opens in its own panel, and you can ask for a specific change without losing the rest of the work.
The underlying pattern
All five of these come down to the same thing: Claude works best as a collaborator you’re actively steering, not a vending machine you put one coin into. The extra ten seconds it takes to say what “good” looks like, or to point at exactly what needs to change, almost always comes back to you as saved time later.