How to Get Genuinely Useful Results from Your AI Assistant

Most disappointing answers from an AI assistant aren’t a model problem. They’re a briefing problem. The assistant only knows what you tell it, so thin input produces thin output. The good news: a few habits fix most of it.

1. Give it the context it can’t see

An assistant doesn’t know your situation unless you say so. Before asking, add the background a smart colleague would need:

• Who this is for: “I’m writing for first-time homebuyers” beats “write about mortgages.”
• Constraints: budget, deadline, tone, length, tools you actually use.
• What you’ve already tried: so it doesn’t suggest the obvious things you’ve ruled out.

One or two sentences of context routinely matter more than clever prompt tricks.

2. Describe the output, not just the topic

“Write about productivity” can produce anything. Instead, specify the shape of what you want back:

• Format: a checklist, a draft email, a comparison table, bullet points
• Length: “about 200 words” or “a one-page outline”
• Tone: “direct and casual” or “formal, for my manager”

If you know what done looks like, say so. You’ll skip a whole round of revisions.

3. Iterate like an editor, not a slot machine

The first draft is a starting point, not a verdict. Treat the assistant like a junior colleague: give specific feedback and ask for a revision.

• “This is too long — cut it to half the length.”
• “Keep the second paragraph, rewrite the rest in plainer language.”
• “Give me two alternatives to the opening line.”

Two or three rounds of this beats re-rolling the same vague prompt five times.

4. Ask it to check its own work

Before you trust an important answer, ask follow-ups:

• “What assumptions did you make in that answer?”
• “What’s the strongest argument against this plan?”
• “Walk me through your reasoning step by step.”

This doesn’t guarantee correctness — you should still verify anything that matters — but it surfaces weak spots fast.

5. Keep the good stuff

When you land on instructions that work well, save them. Many assistants let you set persistent preferences or custom instructions. A short note like “I prefer concise answers with examples” pays off in every future conversation.

The one-sentence version

Brief your assistant the way you’d brief a capable new hire: context up front, a clear picture of the deliverable, and honest feedback on the draft.

None of this requires technical skill. It just requires treating the conversation as a collaboration — which, done well, is exactly what it is.