ChatGPT for Better Work: Context, Not Magic Prompts

Being more productive with ChatGPT doesn’t depend on having a definitive list of prompts saved. It depends on two much more boring things: giving it context and correcting it when it fails. If you copy a “magic prompt” from a social media thread and expect magic, what you’ll usually get is generic, bland text that doesn’t fit what you actually need.

This post is for someone who uses ChatGPT from the chat interface, with no technical knowledge, and wants to actually get value from it in their day-to-day work. You don’t need to know how to code or understand how it works under the hood.

The myth of the perfect prompt

There’s no prompt that works for everything. That “ultimate template” circulating out there was written for a different situation than yours: a different company, a different client, a different goal, a different tone than the one you use.

ChatGPT can’t read your mind. When you give it a vague instruction like “write me a follow-up email,” it fills in the blanks with what’s most likely, which is almost always the most generic option. That’s why the result sounds like a template: because you gave it a template as a request.

Collecting prompts is like collecting recipes without having anything in the fridge. The useful part isn’t the recipe. It’s the ingredients you bring to the table.

Context is almost all the work

The difference between unusable output and something you can actually use almost always comes down to the context you’ve provided. Think about it this way: if you asked a new coworker to write that email, what would you have to tell them first?

They’d need to know who it’s addressed to, what happened in the last meeting, what tone your company uses, and what you’re trying to achieve. ChatGPT needs the same thing. The difference is a coworker would ask you what’s missing, and ChatGPT won’t—it just makes it up.

Good context usually includes:

• Who you are and who you’re writing for (“I’m in sales, this goes to a client who hasn’t signed yet”).
• Your source material: paste the previous email, meeting notes, or the draft you already have in progress.
• Your actual goal, which isn’t always obvious (“I want them to respond, not feel pressured”).
• The format and tone you expect.

It doesn’t need to be polished or organized. Pasting text as-is, typos and all, works better than a fancy abstract instruction.

Iteration is the real skill

The first response is rarely the right one, and that’s fine. Working with ChatGPT is more like directing than requesting. You tell it what’s wrong and ask for another version.

“Too formal, make it friendlier.” Or “too long, cut it down to three sentences and remove the discount part—I can’t promise that yet.” Each correction gives it new information you didn’t have when you started.

This is exactly what you lose when you copy prompts. You only see the corrections hidden behind the “perfect” result in the screenshot. You only see the final photo.

An example: from a vague request to something useful

Compare two ways of asking the same thing.

The vague version: “Write me an email to win back a customer.” What you get is correct and useless, because it could work for anyone.

The version with context: “I’m in sales for an office supply store. This customer used to order every month and hasn’t for three months. Our last contact was about a late shipment—already resolved. I want a short, friendly email that doesn’t sound desperate, reminding them we’re here and inviting them to tell me if they need anything. Don’t mention discounts.”

The second one gives you something you can send with just a couple of tweaks. And notice: there’s no prompt trick here. You just told it what a coworker would need to know.

Where ChatGPT will let you down

Giving it context doesn’t make it infallible. There are things you can’t rely on it for, and it’s worth knowing them ahead of time.

It makes up data. If you ask for a number, a date, or a source you didn’t give it, it sometimes invents them with total confidence. The reason is it’s a program that predicts the most likely text, not a database that looks up facts. When it invents something that sounds real, they call it a “hallucination,” and I explain it in detail in what generative AI is, or why it sometimes makes things up. It’s worth understanding before you trust it with anything important.

Be careful with private data. Pasting customer information, contracts, or any personal data into ChatGPT has implications that aren’t always obvious. Before you do it, it’s worth knowing what a tool like this can and can’t do with your company’s data, which I cover in the guide on what ChatGPT does well and where not to trust it.

How to start today

Pick a task you do every week that bores you: answering similar emails, summarizing meeting notes, writing product descriptions, or preparing a call script. Start with just one.

Next time you do it, instead of writing a short order, tell ChatGPT the full context like you would someone who just started in your position. Then correct the response two or three times until it works for you. Save that conversation: next week, you start from there, not from scratch.

If you prefer a guided path with exercises instead of figuring it out on your own, there’s a short, very practical course on productivity with ChatGPT that teaches exactly this: building your own workflows instead of collecting other people’s prompts.

And when you want to move from using ChatGPT to understanding why it answers the way it does, that’s what our AI without the hype course is about—designed for non-technical people who want to make decisions with confidence.

One new concept every week

Subscribe

Frequently asked questions

How do I use ChatGPT effectively for work?

By giving it context and correcting it. Instead of writing a short order, tell it who you are, what material you have, and what you want to achieve, the way you’d explain it to a new coworker. Then ask for changes in plain language (“shorter,” “more formal,” “remove that part”) until it works for you. That method works for any daily task; a list of other people’s prompts won’t.

Do I need to pay for ChatGPT Plus to be productive with ChatGPT?

Not to start. The free version is plenty to practice what this post covers: giving good context and iterating. Paid plans give you more capable models, but they don’t fix a vague request. If the context is poor, the result will be poor no matter what you pay.

Is there a definitive list of prompts that solves everything?

No, and be suspicious of anyone selling you one. A prompt is tied to a specific context: your company, your client, your goal. What you can reuse is your method of giving context and correcting, which works for any task.

Can I paste my customer data into ChatGPT to work faster?

Carefully. Pasting customer names, emails, or personal data can conflict with privacy and your company’s data protection rules. Ask first what’s allowed in your organization, and when you can, anonymize the information by replacing real names with generic ones to get the same result without exposing anything.

How much context is too much?

You’re more likely to give too little than too much. If you’re unsure, paste the full material and let ChatGPT decide what to use. It’ll ignore what’s extra and make up what’s missing.