In part 1 you sorted your work into three columns: what AI can do, what it can draft, and what stays with you. This lesson turns that list into practice. The aim of the first week isn't to automate anything. It's to learn how an AI assistant behaves with your work, so later decisions are based on experience rather than demos.
Plan on 20 minutes a day. Use any of the main assistants (ChatGPT, Gemini, Claude and Microsoft Copilot). The free plans are enough for this week.
Day 1: set up properly
- Create an account in your own name, not a shared login.
- Switch off model training. On ChatGPT it's Settings → Data controls → "Improve the model for everyone". Other tools have similar settings.
- Install the app on your phone. Most owners do their thinking away from a desk.
- Ask your first question about your own business: "I run a [business] in [city] with [number] staff. Explain how my business works to a new employee in 10 points." Read the answer critically. Where is it right? Where is it generic or wrong? That gap tells you how much context the AI needs from you.
Before you paste anything, read the traffic-light rule in is ChatGPT safe for business data?. In short: public information is fine, internal information needs care, and personal data about customers stays out.
Day 2: a customer message
Take a real customer message from this week that took you time to answer: a complaint, a price negotiation, a detailed question. Remove the name and number. Paste it with a short note about the situation and ask for a reply in your tone.
Then compare. Did it understand the situation? Was it too polite, too long, too stiff? Tell it: "Shorter. Less formal. We don't offer refunds but we can offer a free redo." Notice how much better the second version is. That is the core skill: giving feedback, not accepting the first answer.
Day 3: something internal
Write a staff notice, a job post or a short SOP. Describe it roughly, as you'd tell a manager: "New rule, no phones on the shop floor except for billing. Explain why: customers felt ignored. Polite but firm. In Hindi and English."
Day 4: a long document
Pick something you've been avoiding reading: supplier terms, a lease, a government notice, a long WhatsApp thread exported as text. Ask: "Summarise this. What do I need to do, by when, and what does it cost me? What's unusual?" Then check two or three points against the original. AI summaries are useful and occasionally wrong.
Day 5: a comparison or a decision
Give it two or three quotes, or describe a decision you're weighing. Ask for a comparison table, then ask it to argue against the option you prefer. AI is good at raising points you hadn't considered, and it never gets tired of questions.
Day 6: a repeat task
Pick one task from your "AI can do it" column in part 1. Write down how you do it today, step by step. Ask the AI to do one instance of it. Then ask: "What information would you need from me every time to do this without my help?" The answer is the start of a routine, which is the subject of part 4.
Day 7: review
On one page, write:
| Task | Time before | Time with AI | Quality | Keep? |
|---|
Be honest. Some things will have taken longer because you were learning. Some will have saved a surprising amount of time. Keep the two or three that clearly helped and repeat them next week.
What owners usually learn in week one
- Context is everything. The more you tell it about your business, customers and tone, the better the output.
- The first answer is a draft. Feedback and a second try improve it a lot.
- It's confident even when wrong. Facts, numbers and legal points need checking. See why AI makes things up.
- It's best at the middle column: drafts you finish, not finished work.
Checklist for this lesson
- Account set up in your own name, model training off
- One real customer reply drafted and improved with feedback
- One internal notice or SOP written
- One long document summarised and spot-checked
- One comparison or decision explored
- One repeat task tried, with the information it needs noted
- Review table filled in; two or three keepers chosen
Next in the masterclass: Part 3, giving AI instructions it can follow. For more ideas to try this week, see 12 everyday jobs for ChatGPT in a small business. If you'd like a guided version of this week for your team, that's part of our AI consulting.
