Talking to AI: Prompting Fundamentals

Two parents on the same PTA committee sat down the same night to write the same email: a note asking families to donate to the spring fundraiser. Both used the same AI tool. Maria typed, “Write a fundraiser email.” She got back three paragraphs of warm nothing. “Together, we can make a difference for our children’s future.” She could have written that in her sleep, and it named neither the fundraiser nor the school nor a single reason to give. She closed the tab.

Dana typed this instead: “Write a short email to parents at Lincoln Elementary asking them to donate to our spring book fair, which funds new library books. Warm but not gushy. Mention that last year’s fair bought 400 books, that any amount helps, and that the deadline is April 30. Keep it under 150 words and end with the donation link.” What came back needed two small tweaks and went out that night.

Same tool. Same task. The only thing that changed was the prompt. A prompt is just the message you type to the AI, and it turns out that message is almost everything.

Why this matters to you

You will never touch the inside of an AI. You will never adjust its dials. The only thing you control is what you type into the box, and that is the whole game. Learning to write a good prompt is like learning to ask a good question of a very capable, very literal assistant who knows a great deal about the world and nothing about your life until you tell it.

This is the highest-leverage skill in this entire course. Everything else you will learn later, using AI for planning, for email, for schoolwork help, for small business tasks, sits on top of this one. Get prompting right and every other tool in the course works better. Skip it and you will spend months blaming the AI for answers that were never going to be good, because the question was never good.

The comforting part: it is a small skill. You can learn the core of it in one sitting, and this chapter is that sitting.

The three must-haves: task, context, format

Almost every weak prompt is missing at least one of three things. Almost every strong prompt has all three. Here they are.

Task. What you actually want made. Not “help me with this” but “write a two-paragraph email,” “make a packing list,” “explain this in plain English.” Name the thing.

Context. Everything about your situation the AI cannot see. Who the message is for, what has already happened, what matters to you, what to leave out. The AI starts every conversation knowing nothing about you. The single most common beginner mistake is assuming it can fill in the blanks. It cannot. It will guess, and its guesses are generic because a generic guess is the safest one.

Format. What the finished thing should look like. An email or a bulleted list. A table or a paragraph. Two hundred words or one page. If you do not say, you get the AI’s default shape, which may be nothing like what you needed.

Watch what these three do across different lives.

A soccer coach. Weak: “Write a message to my team’s parents.” Strong: “Write a short group text to the parents of my U10 soccer team letting them know Saturday’s game moved from 9am to 11am because of field maintenance. Friendly, under 4 sentences, remind them to bring water.”

A student. Weak: “Explain photosynthesis.” Strong: “Explain photosynthesis to me like I’m in 8th grade and just failed the quiz. Use one everyday comparison, keep it under 150 words, and end with the one sentence I’d need to get the main question right.”

A volunteer. Weak: “Help me with a thank-you note.” Strong: “Write a warm thank-you note to a local bakery that donated 6 dozen cookies to our animal shelter’s adoption day. Mention that the cookies sold out and helped us cover vet costs. Three sentences, sincere, not over the top.”

Notice that none of the strong versions use fancy words. They just carry more information. That is the entire trick.

Try It Now Pick one real task from your week, something you actually need to write or figure out. First, type it the lazy way, the one-liner you’d naturally reach for. Read what comes back. Now rebuild it with all three must-haves: state the task exactly, add the context only you know (names, dates, what’s already happened, what matters), and say what the output should look like. Paste the new version. Put the two answers side by side. What did the second one know that the first one couldn’t?

Let the AI interview you: the ask-me-first move

Here is the problem with the three must-haves: you do not always know what context matters. You know your situation so well that half of it feels too obvious to mention, and some of it you would never think to include because you did not realize it was relevant.

There is a simple fix. End your prompt with this line:

“Before you answer, ask me any questions you need to make this as good as possible.”

Now the AI does the work of finding the gaps. Instead of guessing, it asks. Dana, writing that book-fair email, might get back: “A few questions first. Is there a suggested donation amount, or truly any amount? Should the tone assume families already know about the book fair, or introduce it? Do you want to mention how the money is used?” Every one of those is a detail that would have made the email better, and every one is a detail Dana had in her head but would not have thought to type.

This one habit will improve your results more than any clever phrasing ever will. It works because it flips the effort. You stop trying to predict what the AI needs and let it tell you.

Try It Now Take the strong prompt you built in the last lab. Add the ask-me-first line to the end and send it again. Answer whatever it asks. Then read its final result. What did it ask about that you hadn’t thought to include? That question is the gap between a decent answer and one you’d actually use.

See how it thinks: show-your-work prompting

You can catch a bad answer after you read it. Better is to catch a bad assumption before it wrecks the answer. Add this to your prompt:

“After your answer, briefly explain your approach and any assumptions you made.”

Now the AI hands you its answer plus a short note: here is how I read your request, here is what I assumed. Maybe it says, “I assumed this email goes to current families, not prospective ones,” and you realize you meant the opposite. You caught it in ten seconds instead of after you hit send. Over time, reading these notes also builds your instinct for how the tool interprets things, which makes your next prompt sharper.

One honest point, because this course will not oversell its own tricks. When the AI explains its approach, it is describing its reasoning in plain language for you. It is not opening a window into the actual machinery inside it and reading you a literal transcript of the computation. What you get is the AI’s account of its approach, generated the same way its answer was. That is not a knock on the technique. The account is genuinely useful and genuinely steerable, because when it names an assumption it made, you can correct that assumption and get a better answer. Just hold it for what it is: a helpful explanation, not a lie detector wired to its brain.

Think About It Think back to the last answer an AI gave you that missed the mark. If it had shown its work first, listed the assumptions it was running on, which wrong assumption would you have spotted and fixed before it ever reached the answer? Odds are the miss was never about the AI’s ability. It was one bad guess you could have corrected in a sentence.

You’re the editor: iterating on drafts

The most useful mental shift in this whole chapter: the first thing the AI gives you is a first draft, not a final answer. You would not expect a human helper to nail a piece of writing in one try with no back-and-forth. The AI is no different. The magic is not in the first response. It is in the second and third.

When the first draft is close but not right, do not throw it out and start over with a new prompt. That wastes the good parts. Instead, give specific feedback, the way an editor marks up a page. “Warmer.” “Cut the second paragraph.” “The ask is buried, make it the first line.” Specific feedback gets you further than rewriting your whole prompt, because the AI already has the context from the first round and just needs to adjust.

One more editor’s move: when you are not sure what you want, ask for choices. “Give me two versions, one short and businesslike, one warm and personal.” Picking between two real options is far easier than describing an ideal you have not pinned down yet. Once you see them, you usually know, and often the best final version borrows a line from each.

Try It Now Go back to the result from your last lab. Don’t start over. Give it one specific piece of feedback, an actual editor’s note: warmer, shorter, cut a part, fix the tone, sharpen the ask. Read the new version. Then give it one more note. Two rounds. Notice how much closer two small corrections get you than a brand-new prompt would have. That back-and-forth is the skill. The first draft was never the point.

What people get wrong here

“There must be magic words.” New users often hunt for the secret phrase, the special way of talking to AI that unlocks better answers. There isn’t one. The difference between Maria’s email and Dana’s was not vocabulary. Dana did not know a password. She simply gave the AI more of what it needed: the school’s name, the reason, the deadline, the tone, the length. Good prompting is not a spell. It is information. Once you stop looking for magic words and start asking “what does this thing not know about my situation yet,” you are most of the way there.

“A bad answer means the AI can’t do it.” You ask for something, the result is weak or generic or wrong, and the natural conclusion is that the tool is not up to the task. Usually that is not what happened. Usually the prompt withheld something the AI needed, and the AI, unable to see your situation, filled the gap with a safe generic guess. The tell is that generic, could-be-anyone quality. When you see it, the fix is almost never a different tool. It is more context, or the ask-me-first line so the AI can pull the missing piece out of you. Blame the prompt before you blame the AI, and you will be right most of the time.

Your move

Download The Prompt Cheat Sheet and keep it where you work. It fits the three must-haves, the ask-me-first move, and the rest of what you just learned onto one page you can glance at until the habits stick.

Then, before you take the quiz, do this once for real. Pick one task you genuinely need done this week, an email, a plan, an explanation, a list. Write it with all three must-haves, task, context, format, and add the ask-me-first line at the end. Answer whatever it asks. Use the result. That one honest round, on something that actually matters to you, will teach you more than reading this chapter twice.

This chapter was developed with AI assistance and reviewed by a human editor. It’s educational, fact-checked where applicable, and may contain minor errors. It’s not a substitute for professional advice.