Free sample lessons

Most people’s first try at ChatGPT goes something like this. You ask a simple question, and you get back a confident, polished answer, with a number, and even a source.

It sounds great. But that source? It may not exist. And the number may not either.

In this lesson you’ll learn what ChatGPT is actually doing when it answers you. Once you see it, you’ll know when to trust it, and when to check.

Here’s the big idea. ChatGPT is a prediction machine. It learned from a huge amount of writing, and it writes by predicting what usually comes next, one small piece at a time.

It’s like the autocomplete on your phone, scaled up enormously. Your phone guesses the next word. ChatGPT keeps guessing, word after word, until it has written a whole email, a summary, or a plan.

That one idea explains almost everything about it. Because it has seen so much writing, it is very good at writing: drafting, rewording, summarizing, explaining, and coming up with ideas.

Ask it to turn a stiff paragraph into a friendly one, and it does it in seconds.

But here’s the catch. It predicts what sounds right. It can search the web, and it can remember things you’ve told it. But it is still writing what usually comes next, so it can be wrong and sound completely sure.

It’s most likely to slip on details that look alike: names, numbers and dates.

Here’s the right way to think about it. ChatGPT is a brilliant first-draft writer, not a source of truth. You bring the facts and the judgment. It brings the speed.

That’s why every lesson in this course uses the same three steps. Brief it: tell it what you need. Refine it: ask a follow-up instead of starting over. And check it: verify what matters before you use it.

Next, you’ll see the first step, the brief, turn a generic answer into an email you can actually send.

Here’s a request almost everyone has typed. Write an email to my landlord about the heating.

And here’s what comes back. Polite, long, full of blanks, and nothing you could send without rewriting it.

The problem isn’t ChatGPT. It’s the brief. In this lesson you’ll learn the three things every good request includes, and watch this email go from generic to ready to send.

Think of ChatGPT as a very capable new colleague on their first day. They write fast, and they’re well read. But they know nothing about you, your landlord, or what has already happened.

Hand a new colleague that one line, and you’d get exactly that email. So give it what a colleague would need. Three things.

First, the goal: what you want to happen. Not just, write an email. I want a repair date this week.

Second, the context: the facts it can’t know. The heating has been off since Monday, and you reported it by phone on Tuesday.

Third, the format: what a good answer looks like. Polite and firm, under a hundred and twenty words. And one more instruction that’s worth its weight in gold: don’t invent any dates.

Put together, it’s four short sentences. Now watch the difference.

Short. Specific. Every fact is yours, and the only blank left is your name. That’s the brief doing its job.

You don’t need to memorize a formula. Before you press enter, ask yourself three questions. Does it know what I want? Does it know what has already happened? Does it know what good looks like?

Try it on the email you’ve been putting off. In the full course, you’ll use the same three parts for reports, summaries, plans and job applications.

You’ve just walked out of a meeting with a page of notes like this. Half sentences, initials, a date with a question mark. And somebody has to send the follow-up.

In this lesson you’ll turn notes like these into a clean action list with owners and dates, and you’ll learn the one instruction that stops ChatGPT from filling the gaps with guesses.

Start with the brief. The goal: an action list. The format: a table with four columns. Action, owner, due date, and open question.

Now the important part. Notes are full of gaps, and ChatGPT’s instinct is to fill a gap with something that sounds plausible. So tell it not to. Only include owners and dates that are actually in the notes, and where something is missing, write: not specified.

Before you paste anything in, take out what’s private: client names, personal details, anything your company wouldn’t want shared. Initials are fine.

Here’s what comes back.

Now look at the gaps. Nobody owns the demo video. The pricing page is waiting on finance. The budget has no sign-off date. In your notes, those were easy to miss. In a table, they jump out, and they’re exactly the questions to raise in your follow-up.

Then refine. Don’t start over. Ask a follow-up: turn this into a short email to the team, with the open questions first.

Last step: check it. Compare it with your notes. Are the owners right? Is Friday really Friday? Did anything private slip in, or anything get invented? It takes thirty seconds, and it’s the difference between a helpful email and an awkward one.

Brief it, refine it, check it. That’s how notes become next steps. This prompt is in the free starter kit, and the full course uses the same method for emails, reports, research and job applications.

It’s Monday morning, there’s an angry customer email, and the quickest thing to do is paste the whole thing into ChatGPT and ask for a reply. Lots of people do exactly that. Before you do, here’s what’s worth knowing.

In this lesson you’ll learn the four kinds of information that should never go in, a thirty-second way to clean up what you paste, and the one document that decides what’s allowed at your job.

Here’s the simplest rule. Treat anything you paste into ChatGPT like an email to someone outside your company. If you wouldn’t send it to a stranger, don’t paste it.

Four things never go in. Passwords and login codes. Personal details about other people: customers, employees, patients or students. Confidential company information, like unreleased numbers or contracts. And anything your company’s AI policy says no to.

But you can still get help with that customer email. Take thirty seconds to clean it up first. Replace the name with a role, remove the account number and date of birth, and keep only what ChatGPT needs to write a good reply.

Now the request has everything it needs, and nothing it doesn’t.

The reply is just as good, and nothing private left your hands.

Two settings are worth a look. In ChatGPT’s data controls, you can choose whether your chats are used to improve the model. And a temporary chat isn’t saved to your chat history.

And at work, one document outranks everything in this lesson: your company’s AI policy. If it says use only the company’s tool, or never paste client data, that’s the rule.

So before you paste, run the quick check. No passwords. No one else’s personal details. Nothing confidential. And allowed by your company’s policy. Then brief it, refine it, and check it, as usual.

Next, you’ll see how to find out what your company’s policy actually allows.

An email arrives from your company. We’ve rolled out an AI assistant; please start using it this week. Maybe it’s Microsoft Copilot. Maybe it’s a business version of ChatGPT. And you’re left wondering what’s different, and what you’re actually allowed to do.

In this lesson you’ll learn what changes when the AI comes from your employer, the four questions to ask about your company’s policy, and why the skills from this course work in every one of these tools.

Here’s the big difference. With a personal ChatGPT account, you sign in yourself and choose your own settings. With a work tool, your company signs the agreement, sets the rules for its data, and may keep records of how the tool is used.

What doesn’t change is how you get good results. The screens look different and the features vary, but every one of these tools works best with a clear brief, a follow-up, and a careful check.

Here’s a brief from earlier in the course, used for a work task. Goal, context, format. It works the same way in a company tool.

Now find your company’s AI policy, or ask your manager, and look for four answers. Which tools are approved? What information can go in? Does AI-written work need a review, or a label? And who do you ask when you’re not sure?

If your company doesn’t have a policy yet, don’t guess. Send your manager a two-line email asking what’s okay. It shows good judgment, and it protects you.

And yes, ChatGPT can help you write that email.

Different tools, different rules, the same three skills. Know what your company’s policy allows, and you can use AI at work with confidence.