Notes: Send messages your colleagues recognize as yours

Start with what Marcus decided. He is a department manager in a prepared example, and he wants to change his team's weekly planning meeting.

After reviewing his week, he settles on four things:

1. Routine status updates are written in the project folder before the Tuesday meeting.
2. The 30-minute meeting is spent on decisions.
3. Each unresolved blocker gets one owner.
4. The new format runs for 3 weeks as a trial, and then he checks whether the meeting still runs over.

Then read the draft. The AI draft on his screen is grammatically smooth, and it announces that the new format will eliminate cross-team coordination delays across the department. Marcus never decided that.

Clearer wording is help I welcome. A changed commitment is different, because the first ordinary project delay after the announcement makes the sender look like someone who over-promised.

Your colleagues plan their week around the commitments in your messages. They also know how you normally write, so a message from a plain-spoken manager that arrives full of stiff buzzwords and grand guarantees makes people tune out or wonder what the polish is hiding.

Adding "make this sound professional" to a request cannot fix this, because the AI does not yet know what your voice sounds like. The next lecture teaches it.

Words like "warm and professional" tell AI very little. Marcus gives it 3 short updates he wrote himself and says out loud what he likes: the decision and its reason in the first paragraph, everyday words instead of corporate jargon, and a clear action for the reader.

He asks the agent to propose recurring writing rules from those examples, with the supporting line quoted beside each rule. The quote is there because an agent can turn one phrase from one old email into a permanent habit.

The agent proposes three rules, and Marcus treats each one differently:

1. Keep. Open with the confirmed decision and its practical reason before the next steps. That matches how he writes on every project.
2. Change. Keep every sentence under 12 words. His messages were short, but a hard limit could cut the reason a colleague needs. He rewrites it as: prefer clear, concise sentences, without removing the reason behind an action or a condition on a decision.
3. Remove. Close announcements with "effective immediately until audit closure". The quote shows it came from a single compliance note 2 months ago.

He also checks that no colleague names or past project dates are left inside the general rules.

Then he tests the rules on a real draft. He applies them to his Tuesday announcement and checks two things: that it sounds like him, and that it still says 3-week trial. When both hold, he asks the agent to save the curated rules as his personal-voice skill.

From then on the skill carries how you write, and each new message only needs the facts of that decision.

Marcus gives the agent his saved personal-voice skill and his confirmed decision record.

The record holds four facts:

1. Routine status updates are posted in the project folder before 9 a.m. on Tuesdays.
2. The 30-minute meeting focuses on decisions and blockers.
3. Each unresolved blocker gets one named follow-up owner.
4. Marcus reviews the format after 3 weeks to see whether meeting spillover dropped.

The same decision creates different work for each reader.

• His 8 team members need to know when the routine starts, where to post their status before 9 a.m., and how to bring a blocker to the meeting.
• His senior sponsor needs the operational reason, how accountability for blockers is protected, and what Marcus will measure at 3 weeks before making the format permanent.

Then he checks both drafts against the record. The sponsor draft says the new format "will keep delivery on schedule", which edges toward a promise he never made. He changes it to "should help us catch blockers earlier, and I will check after 3 weeks whether the meeting still runs over."

The team draft says, "If you are stuck on something, bring it on Tuesday, even if you do not have a fix yet." That keeps early warnings welcome. A draft telling people to exhaust every workaround first would have discouraged them.

Before I approve drafts like these, I check that each reader can tell:

1. What changed.
2. What response I need from them.
3. What I have not decided yet.

Marcus has to present his 3-week trial to the senior sponsor in the weekly delivery review, and he expects questions. Reading his notes word for word will not answer a direct question about why he chose this.

He asks the agent for two things:

1. A short spoken opening built from his decision record: the change, the reason and the 3-week check.
2. 3 plausible questions the sponsor might ask, such as why now, what the team gains, and how he will know the trial worked.

He answers before he reads. He turns on the microphone, as Alex did in the first section, and speaks his answer to the last question. Typing works too.

The question is, "After 3 weeks, how will you know whether the meeting runs over less than it does now?" Marcus talks about how frustrating the long status round has become for his team, and hearing his own answer next to the question shows the gap. He never said how he will measure the change.

When he checks his records, he finds he knows half of the answer. The meeting has been running past its 30 minutes, but he has never counted by how much, so after 3 weeks he would have nothing to compare against.

His revised answer says what he knows and what he still has to check. The meeting has been running over, the new format is meant to bring it back inside 30 minutes, and before the review he will check how long each of the last 4 Tuesday meetings ran.

A good rehearsal does not build a more assertive tone or invent support on the spot. It separates what you know from what you still need to verify, and that is what lets you answer a hard question calmly.

Marcus's messages became trustworthy through changes you could see. Both drafts kept the exact 3-week trial condition, made each reader's requested action clear, and sounded like his direct, considerate voice. In his rehearsal, speaking aloud replaced a story about frustration with a plan to measure the change.

The test I use before my name goes on a draft:

1. Does it preserve my actual commitment?
2. Does it reflect my voice?
3. Does it tell this reader what they need to know, without hiding the tradeoff?

Passing that test is not the end of the exchange. After Marcus sends his announcement, a senior analyst replies that posting by 9 a.m. on Tuesdays will be hard, because she leads an early support call with a client in another time zone.

Asking AI for another polished paragraph that repeats the 9 a.m. rule more firmly would miss what she needs. A finished message cannot supply her experience or solve an obstacle Marcus has not yet heard out.

When the situation moves from announcing a policy to addressing someone's performance, coaching a skill gap or assigning ownership of work, a one-way message is not enough. Preparing for those two-way conversations is the next section.