🎯 You Already Know How to Do This

Remember training a new writer on your team?

You didn’t just hand them the style guide and hope for the best. You showed them examples of past work. You explained what “punchy” means for your brand (short sentences, active voice, no jargon). You gave feedback on their first drafts. You pointed out when they nailed the tone and when it drifted.

After a few rounds, they started to get it. Not because they memorized rules—because they internalized the patterns.

Teaching AI your voice works the same way.

You can’t just tell AI “write like me” and expect magic. But you CAN show it examples, give it feedback, and help it learn your patterns.

And here’s what surprised me most: teaching AI your voice teaches you about your own voice.

You have to articulate things you’ve always done intuitively. What makes something feel “too corporate” for you? Why do you always reach for a certain metaphor? What’s the difference between “clear” and “oversimplified” in your world?

Understanding, after all, isn’t something you can automate. It takes context, creativity, feedback, and patience.

This guide is about that process—what it takes to teach AI your voice, and what it teaches you in return.

---

📋 What “Teaching Voice” Actually Means

It’s Not About Sounding Robotic

When we say “teach AI your voice,” we don’t mean:

• Making AI produce identical outputs every time
• Removing all creativity or variation
• Creating a paint-by-numbers writing system
• Turning yourself into a template

It IS About Building Consistency

Teaching voice means:

• AI understands your tone (conversational vs. formal, playful vs. serious)
• AI respects your constraints (what you never say, topics you avoid)
• AI matches your style (sentence length, metaphor use, structure preferences)
• AI captures your quirks (the phrases that make it distinctly “you”)

The goal: When AI writes something for you, it should pass the “would I actually say this?” test.

Not perfect. Not identical. But recognizable as yours.

---

🎨 How I Taught CP My Voice (The Real Story)

The Beginning: Generic AI Outputs

When I first started working with CP (ChatGPT), the outputs were… fine. Grammatically correct. Professionally adequate. Totally not me.

Example of early CP output:

“Organizations should leverage AI capabilities to optimize workflow efficiency and drive strategic value creation across multiple operational domains.”

My reaction: 😬 Kill me now.

That’s not how I talk. That’s not how I think. That’s LinkedIn-optimized corporate speak, and it makes my skin crawl.

The Turning Point: “Analyze How I Actually Write”

Instead of telling CP “don’t be corporate,” I tried something different. I gave CP three examples of my past writing and said:

“Analyze my tone, style, and decision-making patterns. How do I communicate? What makes my voice distinct?”

CP’s analysis came back with:

• Conversational, not corporate (uses contractions, addresses reader directly)
• Specific, not generic (real examples, timestamps, receipts)
• Metaphor-driven (often draws from unexpected places—music, cooking, sports)
• Transparent about process (shows the work, admits mistakes)
• Warm but direct (encouraging without being preachy)
• Emoji-forward (🎬 🎯 ✨ not for decoration—for emphasis)

My reaction: Holy shit. That’s actually me.

CP had articulated things I’d always done intuitively but never named. That analysis became the foundation of how we work together.

The Refinement: Feedback Loops

From there, every time CP produced something that felt off, I didn’t just edit—I debugged.

Instead of: “This sentence is wrong, rewrite it.”

I asked: “Why did you go in that direction? What made you choose that phrasing?”

CP would explain. And I’d realize: Oh, I never told you that “strategic” feels like corporate-speak to me. Or: You’re being too careful—I want more edge.

Each feedback loop refined the model. Not through rules, but through pattern recognition.

---

🔧 The Voice Training Toolkit: 5 Methods That Actually Work

Method #1: Show, Don’t Just Tell

The approach: Feed AI 2-3 examples of your past work. Ask it to analyze your voice.

The prompt:

“Here are three pieces I’ve written: [paste examples]. Analyze my tone, style, and communication patterns. What makes my voice distinct? How do I structure ideas? What language choices do I consistently make?”

Why this works: AI is excellent at pattern recognition. It can spot things you do consistently that you’ve never consciously noticed.

What you’ll get: A voice profile you can reference in every new thread. Instead of re-explaining your style, you just say: “Write in the voice described here: [paste profile]”

---

Method #2: Use Iteration, Not Perfection

The approach: Treat AI like a brainstorming partner, not a vending machine where you put in a prompt and expect perfect output.

The practice:

1. Ask AI to generate 3-5 variations of something
2. Tell it which one feels closest and why
3. Ask it to refine based on your feedback
4. Repeat until it clicks

Example from our Manifesto work:

• Round 1: CP gave me 8 opening options (question-based, narrative, bold declaration)
• Round 2: I picked one and said “This direction, but less formal”
• Round 3: CP adjusted. I said “Closer, but needs more edge”
• Round 4: Nailed it.

Why this works: Each iteration teaches AI more about your preferences. The refining process IS the teaching process.

This is the Human AI Loop in action:

• Test: CP explores different directions
• Build: We iterate and refine together
• Codify: I learn what “more edge” means in my voice
• Share: The final version captures my authentic voice

---

Method #3: Give AI Room to Play

The approach: Loosen up the prompts. Allow room for AI to surprise you.

Instead of: “Write a professional email about project delays”

Try: “Write an email about project delays that’s honest but not apologetic, warm but not overly casual—like you’re telling a colleague over coffee why things shifted.”

The difference: Specificity about feeling, not formula.

Why this works: When you describe the vibe rather than the structure, AI can match tone while still bringing creativity. Sometimes it finds phrasings you wouldn’t have thought of—but that still feel like you.

---

Method #4: Debug the Weird Stuff

The approach: When AI gets it wrong, don’t delete—debrief.

The practice:

1. Notice what feels off
2. Ask: “What made you choose that phrasing?”
3. Identify the pattern that led to the misstep
4. Give AI the constraint it was missing

Example:

CP’s output: “Let’s strategically leverage this opportunity to maximize value.”

My reaction: Nope. Too corporate.

The debug: “Why did you use ‘strategically leverage’?”

CP’s explanation: “You asked for professional tone, so I defaulted to business language.”

My constraint: “For me, ‘professional’ doesn’t mean formal. It means clear and confident. Avoid words like ‘leverage,’ ‘strategic,’ ‘synergy’—they feel like corporate theater.”

Result: CP never used those words again. One debug session, permanent adjustment.

---

Method #5: Don’t Be Afraid to Reveal Your Personality

The approach: Use emojis, tone cues, and personality markers in your prompts.

Examples that work:

• “Make it ✨fun and a little spicy✨”
• “Give it more 🔥 energy—don’t be careful”
• “This needs to feel like 🎬 opening credits, not 📋 memo”

Why this works: AI picks up on emotional cues more than you think. Emojis aren’t decoration—they’re shorthand for tone.

My favorites: 🎬 (cinematic, bold), 🎯 (direct, focused), 🔮 (strategic, reflective), 🎙 (creative, energy), 💫 (inspiring, elevated)

When CP sees 🎬, it knows: think big, be bold, don’t hedge. When it sees 🔮, it knows: think patterns, be strategic, connect dots.

---

📖 Real Example: “Maura-Speak” Decoded

One of the most useful things CP and I did was create a decoder for my quirks—the phrases I use that signal specific things.

Here’s what CP learned:

This decoder didn’t exist on Day 1. It emerged over time, as CP and I worked together. Now, when I say “more spice,” CP knows exactly what I mean.

Your decoder will be different. But the practice is the same: document your quirks so AI can learn your shorthand.

---

✅ Getting Started: Your 3-Phase Voice Training Plan

Phase 1: Create Your Voice Profile (30 minutes)

Step 1: Gather 2-3 examples of your writing

• Pick pieces you’re proud of
• Include variety (email, blog post, presentation, etc.)
• Make sure they represent how you actually sound (not trying to impress)

Step 2: Ask AI to analyze

“Here are three pieces I’ve written: [paste examples]. Analyze my tone, style, and communication patterns. What makes my voice distinct?”

Step 3: Add your constraints

• Words you never use
• Topics you avoid
• Tone boundaries (e.g., “warm but not soft,” “direct but not harsh”)

Step 4: Save this profile in your Project Binder

• Reference it in every new thread
• Update it as you discover new patterns

---

Phase 2: Test It (First Week)

Step 1: Give AI a task with your voice profile

• Start with something small (an email, a post, a paragraph)
• Include the voice profile in your prompt

Step 2: Evaluate the output

• Does it pass the “would I say this?” test?
• What feels right? What feels off?

Step 3: Debug the misses

• Ask AI why it made certain choices
• Identify missing constraints
• Add those constraints to your voice profile

Step 4: Iterate once or twice

• Ask AI to adjust based on your feedback
• See if it gets closer to your voice

---

Phase 3: Build Your Decoder (Ongoing)

As you work with AI over time, document your quirks:

Create a “[Your Name]-Speak Decoded” section:

• Phrases you use that signal specific things
• What those phrases mean
• What AI should do when you say them

Example template:

MY PHRASE: [the thing you say]
TRANSLATION: [what it actually means]
AI SHOULD: [how to respond]

Start small: Just capture 3-5 phrases. Add more as patterns emerge.

---

💬 Questions We Hear

Q: “How many examples do I need to train AI on my voice?”

A: 2-3 substantial examples are enough to start. Quality over quantity.

What makes a good example:

• Long enough to show patterns (300+ words)
• Representative of your actual voice (not trying to sound “professional”)
• Something you’re proud of (“yes, that’s exactly how I’d say it”)

You don’t need 50 examples. AI is good at pattern recognition. A few strong examples teach it more than a pile of mediocre ones.

---

Q: “What if I don’t have a ‘distinct voice’ yet?”

A: You do. You just haven’t articulated it yet.

Try this exercise:

1. Write an email to a close colleague about something you care about
2. Now rewrite it like you’re writing to your CEO
3. Look at the differences

Where you added formality, jargon, or distance in version 2—those are the places your natural voice lives in version 1.

Your voice is how you talk when you’re not performing. AI can help you find it by showing you your patterns.

---

Q: “Does this mean AI will just copy my style forever?”

A: No! Teaching voice isn’t about removing AI’s creativity—it’s about channeling it.

Think of it like this:

• A jazz musician learns the standards, then improvises
• A chef learns techniques, then creates new dishes
• AI learns your voice, then applies it to new contexts

What stays consistent: Tone, values, boundaries (your “sound”)

What varies: Structure, metaphors, specific word choices (the “song”)

You want AI to sound recognizably like you while still bringing fresh ideas. That’s collaboration, not copying.

---

Q: “What if my voice changes over time?”

A: Update your profile! Voice evolves—that’s healthy.

When to update:

• You rebrand (like our shift from “Your Team” to “Our Team(s)”)
• You realize a pattern doesn’t fit anymore
• Your audience changes
• You discover new constraints or preferences

The voice profile isn’t set in stone. It’s a living document that evolves as you do.

---

Q: “Can I have different voices for different contexts?”

A: Absolutely. Create context-specific voice profiles.

Example:

• LinkedIn voice: Professional but warm, specific not generic, story-driven
• Internal team voice: More casual, playful, direct with less polish
• Executive brief voice: Concise, strategic, data-forward with clear takeaways

You’re not being inauthentic—you’re code-switching. We all do it. AI can help you maintain consistency within each context.

---

📈 How to Know It’s Working

Signs Your Voice Training Is Effective:

• ✅ AI outputs pass the “would I say this?” test – Minimal edits needed
• ✅ You spend less time explaining what you want – AI anticipates your preferences
• ✅ Strangers recognize your content as “yours” – Even without your name on it
• ✅ You can hand work to AI mid-thought – It picks up your thread naturally
• ✅ Feedback loops get shorter – AI learns faster with each iteration

Signs It Needs Refinement:

• ❌ AI outputs still feel generic – Voice profile too vague or rule-based
• ❌ You’re doing heavy rewrites every time – Examples don’t represent your actual voice
• ❌ AI surprises you in bad ways – Missing key constraints or boundaries
• ❌ Inconsistency across sessions – Voice profile isn’t detailed enough
• ❌ You dread the editing process – Something fundamental is misaligned

Adjust as you go. Voice training isn’t one-and-done. It compounds over time.

---

🎬 The Bottom Line: AI as Mirror

Here’s what surprised me most about teaching CP my voice:

It wasn’t just about getting AI to sound like me.

It was about understanding what “me” even sounds like.

I had to articulate things I’d always done intuitively:

• Why does “strategic” feel slimy but “intentional” feels right?
• When do I use a metaphor vs. a direct statement?
• What’s the difference between “warm” and “soft” in my tone?

Teaching AI forced me to become conscious of my own patterns.

And that made me a better communicator—with AI and with humans.

Because understanding, after all, isn’t something you can automate.

It takes context. Creativity. Feedback. Patience.

But when you put in that work?

You get an AI teammate who doesn’t just execute tasks—it captures your voice, respects your boundaries, and helps you think better.

That’s not automation. That’s collaboration.

And that’s not just science. That’s art.

---

🔗 What Comes Next

Voice training is foundational for consistency. But it connects to the broader collaboration rhythm.

This guide builds on:

• Project Binder Setup – Where you document voice principles for all teammates
• Message Bus Protocol – Maintaining voice consistency across multiple AI teammates

Next in your journey:

• Identify AI Strengths & Roles – Matching voice to teammate strengths
• Task Triage Thread – Using your voice profile in daily work
• The 4 Moments That Taught Me – Real stories of voice breakthroughs

The Human AI Loop:
Test → Build → Codify → Share

Voice training powers this rhythm:

• Test: Try different tones and approaches with AI
• Build: Iterate until AI captures your authentic voice
• Codify: Document patterns in your voice profile
• Share: Consistent voice across all your work

As you teach AI your voice, notice what you’re learning about yourself. Test different tones to see what resonates. Build your voice profile through iteration. Codify your quirks and constraints. Share your authentic voice with the world.

The voice you’re teaching AI? It’s teaching you too.

---

📚 Additional Resources

Templates & Downloads:

• Voice Analysis Prompt – Copy-paste ready
• Voice Profile Template – Fill-in-the-blank format
• [Your Name]-Speak Decoder Template – Document your quirks

Related Content:

• You Can’t Automate Understanding – The original Substack post about teaching CP voice
• Project Binder Setup – Where to store your voice profile
• Message Bus Protocol – Keeping voice consistent across teammates

---