Free lesson · Skill · 5 min
Ask better follow-up questions
Lesson 7 of 12 in AI at Work Foundations
The one idea
The first answer is almost never the finished answer. It is a draft to react to. The people who get great results from AI are not better at the opening prompt — they are better at the follow-up.
The weakest follow-up in the world is:
"Try again."
It throws away the good parts along with the bad, and the model has no idea what you actually wanted changed.
Steer with deltas
A delta is a targeted change: keep what works, name exactly what to fix.
"Keep the structure and the first paragraph. Make the tone warmer, cut the third point, and add a clear next step at the end."
Now the model improves the draft instead of gambling on a brand new one.
Three moves that work
- Name what changed: "shorter", "more formal", "lead with the cost" — not "better".
- Protect the good parts: "keep the opening, change only the ending."
- One direction at a time: stacking five contradictory fixes in one breath muddies the result. Make a couple of focused passes.
Try it now
Take any answer the AI gave you that was almost right. Instead of redoing it, send two follow-ups that each name one specific change. Watch how much faster you land on what you wanted.
Your win today
You never have to settle for "almost" again. A follow-up that names what to change gets you there faster than rewriting the prompt from scratch.