Week 0 • Lesson 4 of 5 • 30 mins
Write Your Digital Twin
One document that stops AI writing like it could be anyone.
Write Your Digital Twin
Ask an AI to write something for you without telling it who you are, and it writes for the average of everyone. The average is inoffensive and forgettable.
Your Digital Twin fixes that. It's one document — your role, your audience, how you write, what you can't promise — that you give the AI once. It's the highest-leverage thirty minutes in this course.
1. Why it works
A model with no context has to guess who's asking. It guesses "a generic professional", and writes accordingly.
The Digital Twin replaces the guess with you. Every reply after that starts from your situation rather than from nowhere.
2. Don't start from a blank page
Let the AI draft it from things you've already written, then correct it.
Find three things you wrote that sound most like you — emails, posts, a document — and paste this:
Here are three things I wrote. From them, draft my "Digital Twin"
using the sections below. Infer what you can. Where you genuinely
cannot tell, write [NEEDS INPUT] rather than guessing.
WHO I AM — role, organisation, what I'm accountable for
WHO I TALK TO — audience, what they know, what they care about
HOW I WRITE — sentence length, formality 1-5, phrases I use,
words I never use, how I open and close
CONSTRAINTS — what I can't promise, what I can't share
[PASTE YOUR THREE SAMPLES]
It'll get your audience roughly right and your voice roughly wrong. That's the useful direction to start from.
3. Correct it honestly
Go through the draft and fix it. Two rules:
Be accurate, not aspirational. If you write "I am concise" and you aren't, the output won't sound like you. Describe how you actually write.
Be specific. "Professional but friendly" describes almost everyone. "Short sentences, no exclamation marks, I usually open by answering the question directly" describes you.
Add a "never use" list. Most people have words they hate seeing in their own writing — add them.
4. Install it
Paste these sections into your assistant's custom instructions, replacing the starter version from lesson 2:
- WHO I AM
- WHO I TALK TO
- HOW I WRITE
- CONSTRAINTS
Then add these standing instructions at the end:
- Answer first, explain second.
- Never invent a fact, figure or citation — write [VERIFY] instead.
- If I am wrong, say so in the first sentence.
- Match your length to the question.
5. Test it
Ask it to draft something you'd normally write — a reply to a real email, a short post.
Read it out loud. Does it sound like you? Mark every sentence that doesn't, work out why, and add a line to the twin that would have prevented it.
It usually takes two or three rounds to get right, and then it keeps working for months.
6. What never goes in it
- Passwords, keys, account numbers.
- Client names or anything under an NDA. The twin ends up in conversation history.
- Personal details that aren't about how you work. It describes your work, not your life.
⚠️ Common mistakes
- Writing it from scratch. Draft it from your real writing instead.
- Describing who you'd like to be instead of how you actually write.
- Being vague. "Friendly and professional" changes nothing.
- Never updating it. Revisit it when your role or audience changes.
- Putting confidential information in it.
The full template, with a team version and a review schedule, is in the resources below.
What's next: the lesson people most often skip, and most often regret skipping — measuring where you're starting from.
Resources & Downloads
Hands-on Practicals
Use the extraction prompt on three things you've written, correct what it gets wrong, then paste the WHO I AM, WHO I TALK TO, HOW I WRITE and STANDING INSTRUCTIONS sections into your assistant's custom instructions.
Knowledge Check
What should you avoid putting in your Digital Twin?