ChatGPT Course: 8 Practical Exercises for Real Tasks

👁 114 reads

This short ChatGPT course has eight exercises. Try each one with a real task, not a made-up example: an email, a report, a spreadsheet or sales copy. The sequence follows the training sessions I run for businesses: start with data safety, then move on to writing, research and numbers, and finish by building your own prompt library. Each exercise takes 10-20 minutes.

Exercise 1: Check what happens to your chats

Open ChatGPT settings and check the data controls to see whether your conversations may be used to improve the model. If you use ChatGPT for work, find out whether you have a personal or business account: the terms and data-processing rules differ.

Then list three kinds of information you will never paste into ChatGPT: clients’ and employees’ personal data, company secrets, and information protected by professional confidentiality. Before doing the exercises, replace names and amounts with made-up ones.

Exercise 2: Write a prompt with a role, source material, format and criteria

Many weak answers start with a one-sentence question. Pick a task from your work and write a prompt with four parts:

Translation of the original Polish test prompt: You are a sales representative at a company that sells invoicing software to accounting firms. Below are my notes from a client meeting. Write a follow-up email: no more than 120 words, three bullet points covering what we agreed, and one question at the end. Do not add anything that is not in the notes. [notes]

Compare the result with the answer to a simple “Write a follow-up email” prompt. The instruction “Do not add anything that is not in the notes” matters most: models are prone to filling gaps with invented details.

Exercise 3: Make a style guide instead of asking ChatGPT to “write professionally”

Paste in three pieces of your writing that you think work well and ask ChatGPT to create a style guide: sentence length, words you use, words you avoid, and how you start and end. Edit the guide yourself, because the model may overstate some traits. Add it to every writing prompt from now on.

I use the same approach for whole brands in my training sessions. You can give the guide to a new team member, so your writing no longer sounds as if it came from one machine.

Exercise 4: Research with sources and check for made-up claims

Ask ChatGPT with search to find information about a company you work with, and require a source for every fact. Then open each source and check whether it really says what the model claims. Highlight any facts without supporting evidence.

When I write books on commission, I make a separate fact-checking document: a spreadsheet where every date, name and number has a source. Models are most likely to get dates and names wrong, so I always check those.

Exercise 5: Ask three questions about a long document

Upload a long document, such as a policy, contract or report, and ask three questions: What is it about, in five sentences? What specifically applies to you? Where does it mention deadlines and amounts? For the third question, ask for quotations from the document. You can check a quote in seconds; a summary is harder to verify.

Exercise 6: Get a spreadsheet formula, not a result from chat

Describe your spreadsheet columns and ask for a formula that calculates what you need, along with an explanation of each part. Paste the formula into Excel or Google Sheets and check the result by hand on three rows.

Do not ask the model to calculate a total from a pasted table. It is good at organising and describing information, but it can make addition errors. Let the formula handle the numbers.

Exercise 7: Check AI-written text for familiar patterns

Take a piece of text written with ChatGPT and look for patterns that can make readers suspect AI was involved: an empty summary at the end of a paragraph, lists of three adjectives, “not only … but also”, or a moral at the end of every section. Remove them by hand and compare the before-and-after word counts. The text is usually a few sentences shorter and better for it.

Exercise 8: Build your own library of 10 prompts

Turn the prompts that worked in exercises 2-7 into a file of 10 prompts for tasks you do every week. For each one, note what it is for and what not to paste into it. If you repeat a task every day, turn the prompt into your own GPT. I explain how in my article about GPTs in ChatGPT.

Keep your prompts in a file or shared drive, not in your chat history. You can lose that history when you change accounts, and nobody else can use it.

What to do after this ChatGPT course

These eight exercises teach you how to work with a chat assistant. If you want to earn money from AI services, the next step is to choose one specific service, such as making a video clip, auditing visibility in ChatGPT or building a website for a business. For more on training a whole team to use ChatGPT, see my article on ChatGPT training for employees. Find information about ChatGPT plans and settings in my ChatGPT guide.

Google · Your sourcesQuantslant higher in your Google - one clickAdd to preferred sources →

Frequently asked questions

Do you need a ChatGPT course if ChatGPT is easy to use?

Not to ask a question. A course can help with three things people often get wrong: pasting sensitive data, trusting facts without checking them, and writing prompts without enough context. You can practise all three in this article.

How long does it take to learn ChatGPT?

The eight exercises in this article take 2-3 hours. You will get more confident after a few weeks of using your prompt library for real tasks.

Can you use ChatGPT for financial calculations?

Use it to write formulas and describe data, but not to do the calculations themselves. Let a spreadsheet formula calculate the result, or have a person check it: a model can make mistakes even with simple addition.

Kinetyra