In partnership with
Introducing the first AI-native CRM
Connect your email, and you’ll instantly get a CRM with enriched customer insights and a platform that grows with your business.
With AI at the core, Attio lets you:
Prospect and route leads with research agents
Get real-time insights during customer calls
Build powerful automations for your complex workflows
Join industry leaders like Granola, Taskrabbit, Flatfile and more.
Your AI sounds like AI
Most sales reps use ChatGPT the same way.
They type a prompt. Get generic output. Rewrite it to sound human.
Then wonder why AI isn't saving them time.
The problem isn't the tool. It's that you haven't taught it to sound like you.
Why generic AI output wastes your time
You've seen this happen.
You ask ChatGPT to write a follow-up email. It comes back formal, polite, bloated with corporate speak. You spend 10 minutes editing it down.
You could have written it yourself in 5.
The same thing happens with research summaries. With call prep notes. With prospect outreach.
AI gives you something usable. Not something you'd actually send.
The gap between AI output and your voice costs you the time AI was supposed to save.
The shift that changes everything
Anne from OpenAI said something interesting on a recent panel.
She uses ChatGPT heavily. But she's personalized it so it speaks the way she speaks.
Not just tone. Not just style.
The actual voice.
Her prompts come back sounding like her. No rewrites needed.
That's the unlock.
AI isn't a tool that generates content you edit. It's a tool that generates content you use.
But only if you train it first.
How to personalize ChatGPT to match your voice
This isn't theory. It's a tested process.
Step 1: Feed it examples of your writing
ChatGPT learns from what you show it.
Find 3-5 emails you've written that sound like you. Paste them into ChatGPT.
Say this:
Here are 5 emails I've written. Analyze my writing style. What patterns do you notice?
[paste emails]
ChatGPT will tell you:
Sentence length patterns
Word choice preferences
How you structure ideas
Tone characteristics
This becomes your baseline.
Step 2: Create a custom instruction set
Go to Settings > Personalization > Custom Instructions.
Add this:
When writing for me:
- [sentence length pattern from analysis]
- [tone characteristics from analysis]
- [word choice preferences from analysis]
- Avoid [words or phrases you never use]
When analyzing content:
- Focus on [what matters to your role]
- Skip [what doesn't]
Example:
When writing for me:
- Short sentences, most under 15 words
- Direct tone, no corporate jargon
- Use "you" frequently, avoid "one might" or "it's important to"
- Never use: leverage, synergy, unlock, transform
When analyzing content:
- Focus on what impacts deal velocity
- Skip product features, focus on business outcomes
This tells ChatGPT how to write for you without you specifying it in every prompt.
Step 3: Use meta-prompting to refine your prompts
Most people write bad prompts. Then blame AI for bad output.
Meta-prompting fixes this.
Instead of asking ChatGPT to do the task, ask it to improve your prompt first.
Say this:
I want to [task]. Here's my prompt:
[your prompt]
Refine this prompt to get better output.
ChatGPT will rewrite your prompt. Use the improved version.
This works for any task.
Example:
Your prompt:
Write a follow-up email after a discovery call
Meta-prompted version:
Write a follow-up email after a discovery call where the prospect said their biggest challenge is [X], the impact is [Y], and they need to solve it by [timeline]. Include a proposed next step and a decision-forcing question. Match my writing style: short sentences, direct tone, no corporate jargon.
The output from the second prompt is usable. The first one needs editing.
Step 4: Build a prompt library that sounds like you
Once you've personalized ChatGPT, save your best prompts.
Create a running doc with:
Research prompts
Email prompts
Call prep prompts
Objection handling prompts
Each one includes:
The task
Your style parameters
Expected output format
Example:
RESEARCH PROMPT
Task: Summarize a company's recent news and identify potential pain points
Style: Bullet points, focus on business outcomes, skip fluff
Output: 3 pain points with supporting evidence
[full prompt text]
This becomes your personal AI toolkit.
Copy/paste ready. No guessing. No rewriting.
Why this works
Personalized AI gives you three things:
Speed No more editing output to sound human. It comes back ready to use.
Consistency Your voice stays consistent across all AI-generated content. Prospects don't notice which emails you wrote vs. which AI wrote.
Scalability You can produce more content without diluting your voice. More outreach, more follow-ups, better research.
Anne mentioned this in the panel.
She automates backend tasks so she can be more present in strategic work.
The backend tasks still sound like her. Because she taught AI to write like her.
The meta-prompting advantage
Sam from Cognition made another point worth noting.
He uses multi-step reasoning to build thesis and understand companies deeply.
But even sophisticated AI needs good prompts.
Meta-prompting solves this.
When you ask ChatGPT to refine your prompt, you're essentially coaching yourself.
You learn what makes a good prompt. Your future prompts get better.
It's a feedback loop.
Better prompts → better output → less editing → more time saved.
How to get started this week
Don't try to personalize everything at once.
Pick one workflow.
Start with email.
Collect 3-5 emails you've written that sound like you
Feed them to ChatGPT and ask for style analysis
Add custom instructions based on the analysis
Write one email prompt using meta-prompting
Test the output, refine if needed
Do this for one week.
Then expand to research. Then call prep. Then objection handling.
By the end of the month, your AI sounds like you.
Not like everyone else's AI.
What changes
Kristen from Decagon said she's always optimizing AI to help her be more concise.
Then adds back the personal aspect.
That's the wrong order.
Train AI to be personal first. Then you don't add it back.
It's already there.
Your emails sound like your emails. Your research sounds like your research. Your call prep sounds like your call prep.
AI becomes an extension of you. Not a tool you wrestle with.
That's when the time savings actually happen.
Try this today
Open ChatGPT.
Paste in 3 emails you've written.
Ask: "Analyze my writing style. What patterns do you notice?"
Take the analysis. Add it to your custom instructions.
Then ask ChatGPT to write an email using those instructions.
See the difference.
That's your baseline.
Now refine it until the output needs zero edits.
Once you get there, AI stops being a draft generator.
It becomes your voice at scale.
The Prompts