A new writer's first draft in your voice is usually wrong in the same predictable ways: too formal, too salesy, and full of the exact words you've banned. Not because they're bad, because nobody handed them the system. This kit is the handoff. One hour, five documents, a quiz to confirm it landed, and a review cadence for the first week, built for teams applying the Cohesion pillar.
Before the session: the 5 documents
Send these the day before. If any of them don't exist yet, that's your real first task.
- Voice Charter. Your 25-word "who we are and how we sound" statement.
- Brand Voice JSON. The structured file with personality sliders, banned words, approved phrases, emoji policy, and CTA styles. (If you don't have one, see the Brand Voice JSON Guide.)
- Three gold-standard pieces. Real published content you'd point to and say "this is us." One long-form, one social post, one email. Annotate each with two sentences on why it works.
- One red-pen example. A marked-up draft showing before and after edits, with each edit tagged (tone, clarity, CTA, banned word). This teaches more than any style guide paragraph.
- The banned list. Words and constructions you never use, with one line each on why. "Never 'elevate': it's what every AI-written LinkedIn post says."
The 1-hour session agenda
Step 1: Read the charter out loud together. (5 min) Sounds awkward, works every time. Ask them: "What kind of person does this sound like?" If their answer doesn't match your intent, the charter needs work, not the new hire.
Step 2: Walk the gold-standard pieces. (10 min) For each piece, point at two specific sentences and explain why they're on-voice. Specific sentences, not general praise. "See how this opens with the customer's problem in their words, not our product name."
Step 3: Walk the red-pen example. (10 min) Go edit by edit. This is where the voice stops being abstract. They see what you cut, what you keep, and what a banned word looks like in the wild.
Step 4: Live rewrite together. (15 min) Pull up a generic AI draft (generate one on the spot if needed). Rewrite the first two paragraphs together, narrating your decisions. Then let them rewrite the third paragraph while you watch. Correct out loud, kindly.
Step 5: They write solo. (15 min) Give them a small real task: a social post or a short email. They write it during the session using the Voice Charter and JSON. You review it live, tagging feedback the red-pen way.
Step 6: Set the week-one cadence and hand over the quiz. (5 min) Explain the review schedule below, answer questions, send the quiz.
The 10-question voice quiz
Send after the session. Open-book is fine; the point is retrieval, not memorization. Swap in your own specifics.
- Recite or paraphrase our Voice Charter. What are the three most important words in it?
- Which of these words is banned, and why: [pick three, one banned]?
- Our formality slider is set to [X]. Write one sentence that's too formal for us and fix it.
- Rewrite this sentence in our voice: [insert one genuinely off-voice sentence].
- What's our emoji policy on [channel]?
- Write our default CTA for [your main offer].
- Here are two versions of the same post. Which one is us, and what gave it away? [paste one real, one generic]
- Name two of our approved phrases and use one in a sentence.
- A client asks for something we can't do. Draft the first two sentences of the reply.
- What do you do when AI output uses a banned word: edit the draft, or edit the prompt? (Answer: both, but the prompt edit is the one that counts.)
Scoring: 8 or more right, they're cleared for supervised drafting. 6 to 7, redo steps 3 and 4 of the session. Under 6, the documents are probably unclear; fix them before blaming the person.
The first-week review cadence
| Day | What happens | Time |
|---|---|---|
| Day 1 | They draft one real piece. You red-pen it with tagged edits, not silent rewrites. | 15 min |
| Day 2 | They revise based on tags, then draft a second piece. | 10 min review |
| Day 3 | Compare their day 1 and day 2 drafts together. Name what improved. | 10 min |
| Day 5 | They draft one piece with no guidance. Score it 1 to 5 on charter match. | 10 min |
| Day 7 | Handoff decision: 4+ means they publish with a light check. 3 means one more week of red-pen. Under 3, revisit the documents. | 5 min |
The rule that makes this stick
Every correction goes two places: the draft, and the system. If you fix the same thing twice, it becomes a line in the Voice JSON or the banned list. Onboarding one person properly should leave the system better for the next person. That's the difference between a kit and a favor.