Let Claude interview you, then check and refine a working-context note about you and your work.
Session 1 · Portfolio piece
Teach Claude who you are and what you do
Priya Shah · Meridian Care · version 1 · 11 January
A one-to-two-page working-context note about you and your work.
What I did: I used my assistant on a real piece of my own work, checked the draft line by line and rewrote the two parts it guessed at.
Reflection
The first draft was close but it guessed at two details; I checked them and fixed one. Next time I’ll give it the source material up front so it doesn’t have to guess.
teach-claude-who-you-are.docx · 52 KB
State
accepted by Johnnie on 13 January
Hours earned
Three guided learner hours, counted on 11 January
Johnnie’s comment
“Exactly what I was after: you said what it got wrong and how you fixed it.”
Only Johnnie and the training provider see what you hand in. Your portfolio PDF is yours to keep.
The brief
What you’ll submit
Your final one-to-two-page working-context note, a short reflection and the declaration.
How it’s assessed
Johnnie reviews your work and gives feedback. Resubmit as often as you need.
Hours
Three guided learner hours, counted when you handed it in on 11 January.
Work information only. Please only use information you are allowed to use at work.
Output
A working-context note: a package of information about your job that you paste at the start of any AI chat.
Why it helps. A new chat knows nothing about you, so its answers could be for anyone. With your note pasted in first, it knows the key elements of your job: what you do, who your work is for, what it must never see and what a person must always check. Its answers come back much closer to something you’d actually use.
Steps
Open the app, start a new chat and choose Sonnet as the model. This is a conversation plus a short document, so Sonnet is the right fit; save the most capable model for complex or high-stakes work.
Click Copy prompt to copy the prompt, then paste it in and send it.
Work through the questions with the agent.We highly recommend using dictation for this step. If you’re not sure what information you’re allowed to share with AI, Claude will offer to work it out with you.Tip: the more context you provide about you and your role, the better your working-context note will be.
Once you have your first working-context note, go over it to verify facts and address the outstanding sections at the bottom. Try asking Claude to make the changes for you.
When you are happy with the revision, copy the final note into a document. You’ll use it in the next walkthrough, and later in the course you’ll learn how to stop pasting it altogether.
Paste it into the box below.
Prompt
See the full prompt
I am new to using AI at work and this is a short guided activity. Use plain language, keep your replies short and ask me one question at a time. Do not use or suggest connections to my email, calendar, shared drives or other apps, projects, memory or saved instructions, web search, or anything that runs automatically. Do not save anything from this conversation to memory unless I ask you to. Stick to the one activity described below.
I'd like you to interview me so you can write a context statement about me and my work. I will check it and save it so future chats can start with useful context.
How to run the interview:
- If I have added documents, read them first and tell me briefly what you learned. Do not ask me for information they already contain.
- Ask one short, concrete question at a time, using plain English.
- If my answer is vague, ask one follow-up for a concrete example, then move on.
- If I do not know or do not want to say, write "not known yet" and move on.
- Do not invent anything. Label any guess and ask me to confirm it.
Cover these seven areas, in order:
1. Me and my role: my job title, what I actually spend time on, who I report to and who relies on my work.
2. My team and organisation: what the organisation does, my team's purpose and anything about decision-making that affects my work.
3. My work: recurring tasks and outputs, tools and systems, the people I work with, and current priorities.
4. How I like answers: tone, length, format and anything I dislike.
5. Things you must not do without asking me: for example sending, publishing, changing a shared file, making a decision for me or presenting a guess as fact.
6. Information boundaries: what I may share with an AI assistant, what I must not share, the guidance that applies and who I ask when I'm unsure. If I don't know, offer once to help me work them out by asking a few questions; if I say not now, write "not known yet".
7. Things I am unsure about or still need to find out.
When the interview is complete:
- Write a one-to-two-page context statement using those seven areas as headings and short bullet points.
- End with "Guesses to check" and "Not known yet".
- Ask me what is wrong, missing or too specific, then update it once. Write the final version in the first person, for a future assistant to read, with no notes to me. Keep under "Guesses to check" only the guesses I have not confirmed.
Tip: you can start any chat this way. Asking the AI to interview you first is often easier than working out what to tell it.
What accepted looks like
It’s real work from your own job, with anything private taken out
It does the job: a one-to-two-page working-context note about you and your work
You checked it, and say what the assistant got wrong and what you changed
You · Monday 11 January, 6:30pmHand-in · version 1
Shown as your piece.
AI
AI coach · first read, not the mark · Monday 11 January, 6:32pm
Your note covers all seven areas and says “not known yet” in two places rather than guessing. That is the right call: a future chat can ask you, but it cannot un-learn a guess.
One thing to tighten: the section on what the assistant must never see says “client files” without saying which. One line naming the two or three kinds would cover it.
A question to think about: which part of your week would change most if every chat started with this note?
J
Johnnie · Wednesday 13 January, 10:00amaccepted
Exactly what I was after: you said what it got wrong and how you fixed it.
Every version is kept. It stays accepted until Johnnie marks the new one.
Before you submit. Remove anything you would not share outside your organisation. You can ask Claude to help, then double-check it yourself.
Only Johnnie and the training provider see what you hand in.
Paste your final work.
Draft saved just now
Files (optional)
teach-claude-who-you-are.docx · 52 KB
In your own words or with your AI assistant: what you made, the choices you made and why, and what you’d do differently next time. At least 150 characters.
