Course resource
The Digital Twin Template
One document describing how you work, so AI stops producing output that could have come from anyone.
Write this once. Paste it into any conversation where the output carries your name, or put it in your custom instructions permanently.
Why it works
Without context, a model writes to the average of everything it has read. Your Digital Twin replaces the average with you: your role, your constraints, your audience, your habits. It is the single highest-leverage document in this course, and it takes about thirty minutes.
The template
Fill every section. Vague answers produce vague output.
# DIGITAL TWIN — [YOUR NAME]
Last updated: [DATE]
## WHO I AM
Role: [JOB TITLE AND WHAT IT ACTUALLY INVOLVES]
Organisation: [WHAT IT DOES, HOW BIG]
Industry: [SECTOR AND ANY CONTEXT AN OUTSIDER WOULD MISS]
Experience: [YEARS, AND IN WHAT]
What I am accountable for: [THE THING YOU ARE JUDGED ON]
## WHO I TALK TO
Primary audience: [WHO READS MY WORK]
What they already know: [SO NOTHING PATRONISES THEM]
What they care about: [WHAT MAKES THEM ACT]
What they are sceptical of: [WHAT LOSES THEM]
Secondary audiences: [WHO ELSE SEES IT]
## HOW I WRITE
Sentence length: [SHORT / VARIED / LONG]
Formality (1-5): [N]
I habitually use: [PHRASES, STRUCTURES YOU ACTUALLY USE]
I never use: [WORDS YOU HATE]
Humour: [NONE / DRY / WARM]
How I handle uncertainty: [HEDGE / STATE PLAINLY / ASK]
How I open: [STRAIGHT IN / CONTEXT FIRST / QUESTION]
How I close: [ASK / SUMMARY / NOTHING]
## MY WORK
Recurring tasks: [THE THINGS YOU DO WEEKLY]
Tools I live in: [EMAIL, SHEETS, CRM, WHATEVER]
Typical deliverables: [WHAT YOU PRODUCE]
My deadlines look like: [RHYTHM]
## CONSTRAINTS
I cannot promise: [DISCOUNTS? TIMELINES? OUTCOMES?]
I cannot share: [CONFIDENTIAL AREAS]
Regulatory or compliance context: [IF ANY]
Things my organisation is sensitive about: [TOPICS TO HANDLE CAREFULLY]
## WHAT GOOD LOOKS LIKE
A great piece of work from me: [DESCRIBE, OR PASTE AN EXAMPLE]
A bad one: [DESCRIBE WHAT GOES WRONG]
The mistake I make most: [BE HONEST — AI CAN WATCH FOR IT]
## STANDING INSTRUCTIONS
- Answer first, explain second
- Never invent a fact, figure or citation — write [VERIFY]
- If I am wrong, say so in the first sentence
- Ask one clarifying question when genuinely ambiguous; otherwise proceed
- Match length to the question
Using it
Occasional use — paste at the top of a conversation:
Here is my Digital Twin. Use it for everything in this conversation.
[PASTE]
My task: [TASK]
Permanent use — put the WHO I AM, WHO I TALK TO, HOW I WRITE and STANDING INSTRUCTIONS sections into your custom instructions. Keep CONSTRAINTS there too if you work with sensitive material.
Team use — a shared organisational twin (audience, tone, constraints, things we never promise) plus a personal one per person. The shared part stops five people producing five different house styles.
Building it faster
If the blank page is the problem, extract it from work you have already done:
Here are three things I wrote: [PASTE]
And three emails I sent: [PASTE]
From these, draft my Digital Twin using the template below.
Infer what you can. Where you cannot infer something, write
[NEEDS INPUT] rather than guessing.
[PASTE EMPTY TEMPLATE]
Then correct it. It will get your audience roughly right and your voice roughly wrong, which is the useful direction to start from.
Keeping it honest
The twin decays. Your role changes, your audience changes, your constraints change, and the document quietly stops matching.
- Re-read it when you change role or team
- Update CONSTRAINTS whenever a policy changes
- Add to "the mistake I make most" when you notice a new one
- Review quarterly regardless
| Reviewed | What changed |
|---|---|
What not to put in it
- Passwords, keys, client names under NDA, anything you would not want in a conversation log.
- Aspirational descriptions. If you write "I am concise" and you are not, the output will not sound like you.
- Every detail of your life. The twin is about how you work, not who you are.