Iterating on a result: follow-ups, examples and breaking big tasks into steps

The first answer from an AI tool is rarely the final one, and that is normal. The skill employers look for is not writing a perfect prompt but steering: reading the output, saying exactly what is wrong, and getting the next version closer. This lesson covers the follow-up patterns that work and how to split a task that is too big for one prompt.

Follow-ups that name the problem

"Make it better" gives the tool nothing to act on. Name the gap. Example 1: you asked for a product description and got five paragraphs of adjectives. Follow up with: "Cut this to 80 words. Keep only the three facts a buyer needs: size, material and delivery time. Remove every adjective that is not measurable." The next version is usable because each instruction is checkable.

Useful follow-up shapes:

• Change one thing: "Same content, but written for a customer who is angry, not confused."
• Constrain: "Under 120 words. No bullet points. End with one question."
• Correct a fact: "The refund window is 14 days, not 30. Fix every mention."

Showing instead of describing

An example beats a description of the style. Example 2: you need weekly status notes in the team's format. Paste last week's note and say: "Write this week's note in exactly this structure and tone, using these updates:" followed by your bullet points. If the output copies the example's content instead of its shape, say so: "Use the structure only; the facts come from my notes."

Breaking a big task into steps

A single prompt for "write a 20-page training manual" produces something shallow and uneven. Split it so each step has an output you can check. Example 3:

1. "List the ten topics a new support agent must learn in the first week, one line each." Edit the list.
2. "For topic 3, write the learning goal and five key points." Check the facts.
3. "Expand point 2 into a worked example with a sample customer message and reply."
4. Repeat for each topic, then: "Read these three sections and list any contradictions between them."

Each step is small enough to verify, and errors do not compound. The last step uses the tool as a reviewer, which it does well when the material is in front of it.

Common mistakes and a practice habit

• Starting a new conversation for every follow-up and losing the context.
• Accepting the second draft because it is better than the first, not because it is right.
• Stacking five corrections in one message so the tool addresses two.

Take one real task this week. Write the first prompt, then make yourself do at least three follow-ups, each naming one specific change. Save the first and final versions side by side. After a few rounds you will notice which follow-up shapes you reach for, and the first prompts get better too.