Improve the response one step at a time

Improve the response one step at a time

The first response is a draft you can refine. Instead of repeating the whole prompt, ask for one specific change: shorten the text, add steps, flag uncertain claims, or use a calmer tone. This helps you see what effect each change has.

If the model misunderstood the task, correct the goal first. If information was missing, provide it only if it’s necessary and safe to share. If the answer is too general, ask for an example in a specific situation. Be careful not to keep polishing text that is fundamentally wrong.

When you’re considering a decision, ask for potential downsides or questions that need checking, too. But this doesn’t replace your own review: the model may produce a convincing but incomplete list. Use its suggestions as a starting point for thought, not proof that you’ve considered everything.

If the result repeatedly misses the mark, stop and check whether AI is right for the task. It may lack important human judgment or access to up-to-date information. If so, change your approach or consult the source you need.

Save only the version you’ve checked, and flag any open questions you still can’t answer.

After a few revisions, compare the result with your original goal. If you’ve drifted away from it, revisit the prompt or choose another approach.

How can you usefully revise the model’s first response?

AI in Everyday Life: Use It Effectively and Safely

Continue the training

Enter your email address to continue the training.