Prompting in Plain Language

The most common frustration with AI comes from treating it like a search engine.

You type a few words. You get a response that's technically about your topic but doesn't fit your situation at all. You conclude that AI is overhyped. What actually happened is that you searched when you should have briefed.

Search engines are designed for keywords. You type "patient letter cholesterol" and Google finds relevant pages that other people wrote. The keyword is the input, retrieval is the mechanism.

AI works differently. You're not retrieving something that already exists. You're generating something new — and the quality of what you generate is entirely dependent on the context you give. When you give it a keyword, it generates the most statistically average response to that keyword. When you give it a brief, it generates a response to your specific situation.

That's the entire shift. Not a technical skill. A communication habit.

Brief it like a new assistant

Think about the last time you brought someone new onto a project — an intern, a contractor, a colleague covering for you. You wouldn't hand them a keyword. You'd take five minutes to give them context: who you are, what the situation is, what you need them to produce, what format works best, and ideally an example of something similar done well.

That's exactly how to talk to AI. The investment is identical. The return is dramatically better output.

That single habit change accounts for the majority of the gap between people who find AI genuinely useful and those who don't.

The four parts that make a brief work

A good plain-language ask has four components. You don't need all four every time — a casual question doesn't need a full briefing. But for anything you care about, these four elements are what separate a first draft you can use from one you have to gut and rewrite.

Context is who you are and what the situation is. Not a biography — just enough for the AI to understand your role and the specific circumstances. "I'm a real estate agent in Phoenix. My client is a first-time buyer who just lost their second bid." That's enough. Without context, you get a response that could apply to anyone, which often means it applies to no one.

Task is the specific thing you want. Not "help me with email" but "write a follow-up email after yesterday's showing." The more specific the action, the more specific the output. Vague task = vague result, always.

Format is how you want it back. Three short paragraphs or a bulleted list? Formal or casual? Under 200 words or as long as it takes? If you don't specify, you get whatever the model predicts is most common — which may not be what you need. Specifying format saves editing time.

Example is the highest-leverage part that most people skip. If you have a past email, a past listing, a past patient letter that hit the right tone — paste it in. Even one paragraph of your own writing teaches the AI your voice faster than any description. It is not copying your work; it is learning your style so the output sounds like you, not a generic template.

One rule before anything goes into the box: context means the situation, not the identity. "I'm a cardiologist. A patient just received borderline cholesterol results" gives the AI everything it needs — the patient's name, date of birth, or record number would add nothing but risk. The same applies to a past letter you paste as an example: strip the names and identifiers first. De-identify anything sensitive before you paste it; the honest-limits lesson covers exactly what's safe to share and what isn't.

What this looks like for a real non-technical task

Here's a before-and-after that shows the difference between a keyword ask and a full brief. Same task — a listing description for a property. Same AI tool.

The vague ask took ten seconds to type. The specific brief took forty. The specific brief saved thirty minutes of editing.

That math holds up across almost every professional task. The effort you put into the brief comes back to you multiplied in the output.

How to continue the conversation

The brief is the start, not the entire conversation.

When the first response isn't quite right, you don't start over. You direct it. "That's too formal — bring it down one level." "The second paragraph is perfect. Rewrite the first to match that tone." "Add a sentence about the renovation timeline." You talk to it the same way you'd redirect a person who gave you a good first attempt but missed one thing.

This is the part that feels counterintuitive at first. You're not just sending prompts — you're having a directed conversation with a capable assistant. The context you gave in the brief is still there. You don't have to restate it. You just direct the next iteration.

What to do when you're stuck on the brief

If you're not sure how to start, try finishing these sentences before you open the AI tool:

• "I'm a [role] and the situation is..."
• "I need you to [specific action]..."
• "I want it to be [format/length/tone]..."
• "Here's an example of what I mean: [paste]..."

You don't need all four every time. Even one good context sentence is better than nothing. Start there and build up.

Before you put any of this into practice, the next lesson helps you pick the AI tool you'll brief in — there are a handful of mainstream options, and the choice is simpler than it looks. After that, the lesson on the honest limits of AI covers what to do once you have a draft: how to fact-check what matters and make the output actually sound like you before it leaves your hands.

For professionals ready to go deeper — building workflows, automating repetitive tasks, and developing the full skill set — jeremyknox.ai maps out what the full AI operating system looks like beyond the basics.