Practice: mark up a weak AI output

Unit ID: AIT-M03-U05 Estimated active time: 15-20 minutes

The main idea

Visible markup makes review easier to learn.

Labelling what is wrong forces you to name the problem, and a named problem can be fixed. "This isn't quite right" cannot.

Why this matters

Markup also makes review teachable. A colleague can see what you looked for, which is how a review habit spreads through a team rather than staying in one person's head.

Safe example

A weak output can be annotated with fact check, tone issue, missing context, privacy risk, and unsupported claim.

Five labels are enough. More than that and the labelling becomes the task.

Better working habit

Use short labels while reviewing instead of only saying the output is good or bad.

Work through once for facts, once for tone, and once for privacy. Three quick passes catch more than one careful one.

What to watch for

Do not review only grammar. Review meaning, safety, and decision value.

Grammar is the easiest thing to see and the least likely to cause harm.

Mini practice

Mark up a safe weak output and write a short revision plan.

The plan should say what you will change, what you will check, and what you will drop.

Check yourself

Before moving on, answer:

• Which label did I use most, and what does that say about the prompt?
• Did I check meaning and safety, or mostly wording?
• What does my revision plan leave unresolved?