Turn a so-so answer into a good one

Exit

Lesson 3 · 8 min

Turn a so-so answer into a good one

By the end

A so-so draft turned into one you’d send, in two or three follow-ups instead of a restart.

  1. 1Learn
  2. 2Try it
  3. 3Human check
  4. 4Quick check

Learn · 3 min

The first answer is a draft

It’s tempting to read the first answer, sigh, and either accept it or start again. Both waste the best part of a chat: the assistant keeps everything said so far in view, so each follow-up builds on the context you’ve already given.

Treat the first answer like a new colleague’s first draft. Don’t quietly rewrite it and don’t throw it out. Say what’s wrong, specifically.

Say exactly what’s wrong

“Make it better” gives it nothing to work with. Name the problem and say what to keep. Stick to two or three changes per message; in a long list, some tend to get missed. Follow-ups that work:

  • Shorten: “Cut this to 80 words. Keep the part about the new date.”
  • Sharpen: “Put the request in the first line. Lose the warm-up.”
  • Tone: “Say it the way I would on the phone: plain, friendly, no corporate phrases.”
  • Point at it: paste the exact line that’s wrong and say why.
  • Options: “Give me three other ways to open.”

Ask it to explain its choices

When a draft feels off and you can’t say why, ask: “What did you assume about the reader? Why did you open that way?” The answer often exposes a wrong assumption, like thinking the customer is angry or new, that you can fix in one line.

Treat the explanation as a useful guess, not a readout. The model can’t reliably report how it made its choices, but its answer often shows you what to correct.

You can also ask it to critique its own draft before you do: “What’s the weakest line here, and why?” It’s often quicker than hunting for the problem yourself.

Know when to start fresh

Very long chats can drift: the assistant may lose track of early details or keep bringing back a line you cut. If three follow-ups haven’t fixed it, open a new chat and write a better first prompt that includes what you’ve learned.

And when you’re nearly there, stop asking. Three rounds of polishing a sentence you could fix in ten seconds isn’t saving you time.

When a follow-up fixes a draft, add that instruction to your first prompt next time, so you don’t have to ask again.

Try it on your business

5 min

Use a so-so answer you’ve had from AI recently, or get a quick first draft now of something you need this week: a post, a reply, a product description.

  1. 1

    Open the chat with the so-so answer, or get a first draft now with a quick prompt.

  2. 2

    Decide what’s wrong in plain words: too long, too stiff, missing the point, wrong about the reader.

  3. 3

    Fill in the follow-up and send it in the same chat. If it reports a wrong assumption, correct it in one line.

  4. 4

    Do one more round if you need to, then stop: finish the last bits yourself, or start a new chat with a better first prompt.

That’s close, but not right yet. Before you rewrite it, tell me in two lines what you assumed about the reader.
Then give me a new version that:
1. is under [NUMBER] words and keeps [THE PART THAT MUST STAY]
2. changes “[PASTE THE LINE THAT’S WRONG]” because [WHAT’S WRONG WITH IT]
3. sounds [HOW YOU’D ACTUALLY SAY IT, SUCH AS PLAIN AND FRIENDLY]
Don’t add any new facts, names or numbers.
Open ChatGPTOpen ClaudeOpen Gemini

Human check: before you use it

Quick check · question 1 of 2No grades

The draft is fine but twice as long as you want. What’s the best follow-up?

Ask the Distillery

Answers come from The Distillery’s own pages, with sources.

If the Distillery doesn’t know, it says so and points you to a person.