Practical AI Tips for Getting Better Results
Most disappointing AI output comes from how the request was made, not from the tool being incapable. A few small habits make the difference between a vague, generic answer and something you can actually use. These are the ones I rely on most.
1. Give Context Before the Question
An assistant only knows what you tell it. Before asking, say who the output is for, what you're working on, and what constraints apply:
Weak: "Write a function to parse dates."
Better: "Write a TypeScript function that parses ISO 8601 date strings from
an API response. We already use date-fns, and invalid input should
return null instead of throwing."
The second version removes a dozen guesses the assistant would otherwise make for you.
2. Be Specific About the Output
Say what shape you want back: a table, five bullet points, a diff, under 150 words, in a friendly tone. If you don't specify a format, you get whatever the assistant defaults to, which is rarely what you wanted.
3. Show an Example
When the format or style matters, paste one example of what good looks like. A single sample communicates tone, structure and level of detail better than a paragraph of description.
4. Break Big Tasks into Steps
Asking for an entire feature, report or essay in one go produces shallow results. Instead, ask for an outline first, agree on it, then fill in one section at a time. Smaller steps are easier to correct before mistakes pile up.
5. Iterate Instead of Starting Over
The first answer is a draft. Reply with what to change: "shorter", "assume the reader is a beginner", "this misses the edge case where the list is empty". Targeted follow-ups usually beat rewriting the whole prompt.
6. Verify What Matters
AI can be confidently wrong. Treat output like a colleague's first draft:
- Run code and tests before trusting it.
- Check facts, numbers, names and citations against a real source.
- Be extra careful in security, legal, medical and financial topics.
The more costly a mistake would be, the more checking the output deserves.
7. Know When Not to Use It
If a task takes thirty seconds by hand, or needs private information you shouldn't share, or depends on judgment only you can make, skip the assistant. Never paste secrets, credentials or customer data into a tool you haven't vetted.
Summary
Good results come from clear context, a specific request, small steps and honest verification. None of this is magic, and it's the same discipline that makes working with people go well. Treat the assistant as a fast, tireless collaborator who needs good direction, and you'll get far more out of it.

