Portmint Lighthouse

Say What You Actually Want

Last time we found out why vague questions get vague answers. Today's fix is simpler than it sounds: say the thing you actually want.

Most people ask the AI a half-question. "Can you help with my resume?" "Something about emails." These aren't really requests — they're throat-clearing. The AI hears them and guesses, because you've given it nothing firm to aim at.

The plain idea: state your real goal out loud. What's the task, who is it for, and what would make you nod and say "yes, that's it."

Three things make a goal clear

Think of ordering at a café. "I'll have a coffee" gets you something, but maybe not what you wanted. "A large oat-milk latte, extra hot, to go" gets you exactly your morning. A few more words, far less disappointment — because you named the size, the details, and how it'll be used.

A good request to an AI works the same way. It carries three small parts:

The task — the actual verb. Write, summarize, compare, fix, plan. Not "help with," which could mean anything.

The audience — who reads or uses the result. A note for your boss reads differently than one for your sister. The AI can't see the person on the other end unless you point at them.

The good result — what "done well" looks like to you. Short and friendly? Formal and thorough? Three options to choose from? You know what you're hoping for. Say it.

Watch a half-question become a real one

Half-question: "Can you help me with a thank-you message?"

Real request: "Write a short thank-you email to a customer who left us a kind review. Warm but professional, about four sentences, no salesy language."

Notice nothing fancy happened. No magic words, no tricks. You just told the truth about what you needed — the task (write a thank-you email), the audience (a customer), and the shape of a good result (short, warm, not salesy). The AI now has a target instead of a fog.

This is the same instinct behind a well-built business assistant. When Portmint trains an AI for a company, half the work is teaching it exactly what that business wants a good answer to look like. You can do a smaller version of that in every message you send — just by being plain.

One honest note: stating your goal clearly won't make the AI read your mind about facts it doesn't have. That's the next part of the voyage. For now, we're only making sure the aim is true.

Your turn

Take a request you'd normally type as a quick half-question. Rewrite it with all three parts spelled out: the task (the verb), the audience (who it's for), and what a good result looks like. Read it back — would a stranger know exactly what you wanted?

Next, we'll feed it the facts it can't guess — the context that turns a clear aim into a right answer. 🔦

Stuck or curious?

Ask Pip about this lesson — tap the porthole bottom-right.