← All guides
ORGANISATION

The context file: 4 documents your AI assistant reads before every task

You have four documents written once. After that, your AI assistant knows your client, your offer, your voice and your proof.
DURATION60 minOnce. After that, 5 minutes a month to update
LEVELMediumYou answer questions, then store four documents in a project
WHAT YOU NEEDA Claude or ChatGPT accountThe free version is enough to write the documents. A project (Claude) or custom instructions (ChatGPT) to keep them

What is this guide about?

Without context, your AI assistant answers you like anyone else. With four short documents it reads before every task (your ideal client, your offer, your voice, your proof), its answers sound like you and speak to your target, from the first try.

What you will have at the end

  • 4 short documents: your ideal client, your offer, your voice, your proof.
  • An opening message to paste before each task.
  • A small update ritual so the file stays true.
SalesOffer & pricingAcquisitionContentOrganisationMindset
The bigger the red area, the more this guide makes you work on that skill.

Before you start: gather your material (15 minutes)

  • Texts of yours: 3 to 5 pieces you wrote or said (posts, emails, transcripts). They will be used to describe your voice.
  • Your offer: what is included, the price, who it is for.
  • Real sentences from your target: messages, calls, comments. Delete first names.
  • Your proof: verified results, testimonials, screenshots. If you have none, the file will note it.

The message to copy

You are going to build my context file: 4 short documents that an AI assistant will read before every task so it answers like me and for my target. Talk to me in simple words, one question at a time.

Rules: invent nothing. Each document fits on one page at most. If information is missing, write (to be filled in by me) and tell me how to find it. Delete first names and any personal data.

The 4 documents:
1. MY IDEAL CLIENT: who they are, their situation, their 3 main difficulties ranked, what they've already tried, what they want, the exact words they use, where I find them.
2. MY OFFER: who it is for, what you get, what is included and excluded, price, duration, how it goes, what I don't promise.
3. MY VOICE: how I speak (tone, sentence length, words I like, words I refuse, examples from my texts), with 5 typical sentences and 5 “never me” sentences.
4. MY PROOF: verified results, testimonials with permission, steps of my journey, figures I can quote. Nothing unverified.

Step 1. Ask me questions for document 1, one by one, then write it. Ask me to validate before moving on.
Step 2. Same for document 2.
Step 3. For document 3, ask me to paste 3 to 5 of my own texts, analyse them and write my voice sheet. Quote exact sentences.
Step 4. Same for document 4.
Step 5. Write an “opening message” of 6 lines to paste before each task: “You work for me. First read these 4 documents. Say nothing my proof doesn't allow. Write in my voice. Speak to this client. Respect my offer.”
Step 6. Give me the list of the 4 documents, each in a separately copyable block, and a once-a-month update ritual (3 questions to ask me to check everything is still true).

How it works

  1. Open a new conversation in your assistant, paste the message above and answer its questions.
  2. Save each of the 4 documents:
    • Claude: create a project, add the 4 documents as project knowledge.
    • ChatGPT: create a custom GPT (or paste the essentials into custom instructions).
  3. Before a task, paste the opening message (or work inside the project), then give your request.

To get the most out of it

  • Short beats long. An assistant reads 4 clear pages better than 40 mixed ones.
  • Only put in what’s true. If a proof is approximate, the file will make all your texts lie.
  • Read your voice out loud. If a typical sentence doesn’t sound like you, replace it.
  • Update monthly. A changed offer or a new proof should go into the file immediately.
  • Don’t share client data. Keep the file free of first names and personal information.

Frequently asked questions

Is it the same as “connect Claude to my tools”?

No. Connecting your tools gives access to your emails and calendar. The context file explains who you are and who you talk to. The two complement each other.

I have no proof or clients. Is it useful?

Yes. The “proof” document will note “none for now” and tell you what to collect. The assistant will then avoid promising what you can’t prove.

Do I have to start over in each new conversation?

No, if you use a project (Claude) or a custom GPT (ChatGPT). Otherwise, paste the opening message and the documents at the start of each conversation.

Ready to stop re-explaining everything to your assistant?