Header logo
Copied to clipboard
Creative Authority
Five Star Live · AI Series

Make your AI
sound like you.

Three ways to stop getting the average agent's email back, and start getting yours.

Scroll to begin
Matt Goldman · Fellow Five Star Agent Beginner → Reusable → Operating System
The Premise

Voice is not just language.
It is judgment expressed through language.

Left alone, AI predicts the average. Ask it for an email and it writes the email the average agent would write, because that is what it was trained on.

It does not sound like you because you have not given it a specific human pattern to work from. That is a structural problem, not a prompting problem.

Today is three ways to give it your pattern. Each one goes deeper than the last.

The Three Methods

From a one-off fix to a system that holds.

01

Mimic a Sample

Hand it one thing you wrote and ask it to match your patterns for a single email.

Beginner · The one-off
02

Reusable Voice Reference

Give it several examples and have it derive rules you can reuse across every draft.

Intermediate · Repeatable
03

The Operating System

Mission, decision rules, and canonical voice working together so the output is authored, not generated.

Full build · Durable
01
Beginner · The one-off email

Give it a sample. Ask it to mimic.

The Prompt
Here is an email I wrote that sounds like me:
[PASTE EMAIL]

Study the sentence length, word choice, structure, pacing, and how I give advice.

Now draft an email about [TOPIC] using those characteristics. Do not copy sentences or facts from the example. Use it only as a voice reference.

Before writing, ask me any questions you need about the audience, situation, and desired outcome.
Good for
The single email you need to sound right, right now.
The ceiling
It copies the surface. Nothing carries to the next chat.
02
Intermediate · Reusable

Several examples. Let it find the pattern.

The Prompt
I'm going to give you several examples of my communication. Identify the recurring patterns in:
- How I begin
- How I organize information
- How I explain trade-offs
- Sentence length and rhythm
- Words and phrases I naturally use
- Language I avoid
- How I establish authority
- How my written and spoken voices differ

Separate your conclusions into:
1. Written voice rules
2. Spoken voice rules
3. Language to avoid
4. Questions you should ask before drafting

Every rule must be supported by evidence from my examples. Do not invent personality traits or make my voice more polished than it is.

Keep your full output to 8,000 characters or fewer so it fits in a saved instructions field.
Good for
A reference you paste into any chat or save in a project.
The ceiling
Rules without anchor examples degrade. It drifts back to polished over time.
03
Full build · The operating system

Two documents. Three layers.

Your Mission Statement and your Canonical Voice Reference build something the first two methods cannot: a system that knows how you decide, not just how you sound.

MissionWhat I believe and what matters. Who I serve, the outcomes I prioritize, what I will never recommend, what stays human.
Decision rulesHow I evaluate choices and boundaries. Enforceable IF / THEN logic and when the AI must stop and ask.
Voice referenceHow those decisions sound out loud. Written and spoken rules, objection patterns, prohibited language, approved anchor examples.
How a Request Flows Through It

Asked to write an email, it does not start with word choice.

It moves through the sequence first. Only then does it draft.

  1. 1Who is receiving this?
  2. 2What is happening?
  3. 3What outcome matters?
  4. 4What would this person genuinely advise?
  5. 5What trade-offs need to be explained?
  6. 6Written or spoken channel?
  7. 7What language or behavior is prohibited?
  8. 8What information is missing?
  9. Only then should it draft.
03
The system in motion

After it drafts, it checks its own work.

Quality-Control Prompt
Review this draft against my Mission Statement, Decision Rules, and Canonical Voice Reference. Check:
- Does the advice reflect how I actually think?
- Did you preserve my intent?
- Does this fit my written or spoken voice?
- Did you add hype, urgency, assumptions, or invented specificity?
- Are there phrases I would not naturally use?
- Is anything missing that should be clarified before this is sent?

Revise only where the draft violates one of those standards. Do not polish it merely to make it sound more impressive.
The pointThe check is not to make it fancier. It is to catch anything that is not yours.
Why This One Is Different

The first method imitates the surface. The third teaches it the connection.

What I believeMission
→
How I decideDecision rules
→
How I communicateVoice

That chain is what makes the output feel authored instead of generated. Method one borrows your voice for an hour. Method three gives the tool your judgment to keep.

Where To Start

Start with one email today.
Build the system this month.

Use Method 1 tonight on the next email you have to send. Feel the difference between the average agent's draft and yours.

Then, when you are ready to stop rebuilding your voice in every new chat, we build the operating system together.

Authorship is the competitive advantage.