Set up and tell it who you are
Write three lines about your work once, and put them where the AI reads them every time.
Prevents mistakes like "The same request, two replies".
I want a short note about my work that I can paste above my requests to you. My work in one line: [your job and the kind of work you do most] Ask me three questions, one at a time, and wait for each answer: what I do in a normal week, who reads what I write, and how I like things written (length, format, words I avoid). If I answer with a word like "professional" or "friendly", ask me for an example sentence instead. Then write the note in plain text, under 120 words, in three short parts: Who I am. Who I write for. How I want things written. Use only what I told you. If I said little about something, keep that part short. Don't ask for client names, money figures or personal details about other people. If I mention any, leave them out and tell me you did.
Write three lines about your work once, and put them where the AI reads them every time.
Imagine you keep the accounts for a few small shops. One client's monthly summary will be two days late. You ask AI:
Draft two lines telling a client their September GST summary will come on Thursday, not Tuesday. Two suppliers sent their bills late.
It gives you:
Dear Valued Client, we regret to inform you that the delivery of your September GST summary has been delayed until Thursday. We sincerely apologise for any inconvenience this may cause and appreciate your patience.
The facts are right. You gave them. Everything else is a guess: a formal letter, to a stranger, from a company.
Now paste three lines above the same request:
I keep accounts for eleven small shops in Nashik and file their GST. I write to shop owners on WhatsApp. They read between customers. Two or three sentences. Main point first. No "Dear Sir/Madam". Plain words.
Send it again, word for word. This time you get:
Your September GST summary will reach you on Thursday, not Tuesday. Two suppliers sent their bills late, so I'm waiting on those.
You could send that. The request didn't change. The three lines did the work.
The first reply wasn't careless. The AI didn't know who you were, who would read the message, or how you write. So it wrote for nobody in particular.
A new colleague would have asked you the same three things on their first morning. The difference is that a colleague asks once. The AI starts every new chat knowing nothing about you.
You can retype the answers each time. Or you can write them down once.
5 to 10 minutes. You'll write three lines about your work. Pasted above any request, they replace the explaining you'd otherwise retype. You'll test them before you finish.
You need one AI tool open. A free account is fine.
Here's a filled example:
| Question | Example |
|---|---|
| Who are you, and what do you do? | I run a printing shop in Pune. We print wedding cards and shop signboards. |
| Who reads what you write? | Customers, on WhatsApp. Most order once and worry about the date. |
| How do you want things written? | Two or three sentences, friendly, no "Dear Sir/Madam". Never a date I haven't given. |
If the two replies look the same, your lines are too general. Add who reads your messages and how long you like them.
If your tool has standing instructions, paste the lines there once you're happy with them. If you can't find the setting, pasting them above a request works the same.
Your answers save automatically in this browser. "More details" has room for what you're working on, what you always delete, and what you've decided not to paste.
About you
Three lines about your work. Paste them above a request and the reply already knows who it's for.
For example: I run a printing shop in Pune. We print wedding cards and shop signboards.
For example: customers, on WhatsApp. Most order once and worry about the date.
Defaults, not adjectives. For example: two or three sentences, friendly, no "Dear Sir/Madam".
More details (optional)
The two or three live things. Put a date on it.
The lines you always delete, such as "I hope this message finds you well".
One line, after asking what happens if it ends up somewhere you didn't choose.
Send one request with your three lines and once without. Can you point to what changed?
Can you say who reads what you write, in one line, without using the word "stakeholders"?
Before you paste something sensitive, what is the one question you ask?
Your three lines stay saved in this browser. Lesson 3 and Lesson 4 use them, and Lesson 12 checks they're still true.
Next, you'll take one request you really send and rewrite it so it works the first time.
For now, remember:
Write it once. Put it where the AI reads it every time.