You generate a draft with AI, read it, and start editing. An hour later you've rewritten half of it, and the AI "saved" you exactly zero time. That hour isn't wasted, though. Every edit you just made is a rule you haven't written down yet.
Most people close the doc and move on. That's the mistake. The edits you made this time are the same edits you'll make again next time, unless you write down what you learned. The fix is a five-minute habit called the Red-Pen process, and it compounds every time you use it.
Every Edit Is an Instruction in Disguise
Change "We provide solutions" to "We fix problems," and you're not just editing, you're teaching: prefer direct language over corporate speak.
Delete a paragraph of throat-clearing and jump straight to the point, and you're saying: lead with value, not setup.
Add a casual aside or a bit of humor, and you're saying: we're humans talking to humans, not a brand reading from a script.
The question is whether you're capturing these lessons or re-teaching them every single draft.
The Red-Pen Process, Step by Step
Step 1: Edit Mindfully
Next time you edit AI output, slow down before you change something and ask why. Tag each edit with a category:
- Tone: too formal, too casual, wrong emotion
- Clarity: confusing, wordy, missing context
- Voice: doesn't sound like us, wrong word choice
- Structure: wrong order, missing section, too long
Step 2: Extract the Pattern
After a few editing sessions, patterns show up on their own:
- "I always cut the first paragraph."
- "I always change passive voice to active."
- "I always add a specific example."
A pattern that repeats three times isn't a preference anymore. It's a rule. The tricky part is that most people notice these patterns subconsciously and never write them down, so the same three edits keep getting made, draft after draft, forever. Writing the pattern down is what breaks the cycle.
Step 3: Write the Rule
Turn each pattern into an instruction clear enough that someone else could follow it without you in the room.
| Pattern | Rule |
|---|---|
| "I cut the first paragraph" | "Start with the main point, not context." |
| "I change passive to active" | "Use active voice. Say who does what." |
| "I add examples" | "Include at least one specific example per section." |
The Red-Pen Feedback Template gives you the full worksheet for running this on your own drafts.
Step 4: Feed It Back
Add the rule to your prompts. Add it to your Brand Voice JSON. Share it with your team so the next person editing AI output isn't relearning the same lesson from scratch.
What the Compound Effect Looks Like
Each editing session sharpens your prompts. Each sharper prompt means less editing next time. This is the same loop the Feedback framework is built around: every correction is data, and data compounds if you actually record it.
After a month of doing this consistently, you'll have:
- A written list of voice rules pulled straight from your own editing habits, not guessed at
- AI output that needs noticeably less cleanup per draft
- A shared vocabulary your team can use to describe "what we sound like" without arguing about adjectives
A Quick Way to Start Today
You don't need a system before you begin. Grab your next AI draft and, before you touch a word, open a blank note. Every time you change something, write one line: what you changed and why. By the end of the draft you'll have five to ten raw entries. That list is the seed of your rule set.
Do this for a week and the categories from Step 1 start sorting themselves. Most people find that two or three categories account for the majority of their edits, and that concentration is a signal about where their prompts are weakest right now. Fix those two or three categories first, and the rest of your editing time shrinks along with them.
Next Step
Run your next AI draft through the Red-Pen Feedback Template and turn today's edits into tomorrow's better first draft.