Prompt Engineering for Non-Techies

Chapter 1: Why Your AI Outputs Suck (And It's Not the AI)

You've probably been disappointed by AI at least once. Maybe several times. You asked ChatGPT to write a promotional email and got something generic and hollow—the kind of thing that would make your sales team groan. Or you prompted Claude to brainstorm campaign ideas and got back five variations of the same safe, uninspired concept. Or you tried to get Gemini to analyze your customer feedback and it gave you a summary so vague it could apply to any business selling anything.

Your reaction was probably to assume the AI was limited. That it wasn't smart enough to understand nuance. That these tools were overhyped.

Here's the hard truth: the AI isn't the problem. You are. And I don't mean that as an insult—I mean it as a diagnosis, because it means you can actually fix this.

The outputs you're getting aren't mediocre because the AI is mediocre. They're mediocre because your prompts are mediocre. And unlike the AI's capabilities, which you can't change, your prompts are entirely within your control.

The Vague Request Problem

Let me show you what I mean with a real example. You're a marketing manager at a mid-size SaaS company. You need to write an email announcing a new feature. You're busy, so you open ChatGPT and type:

Write an email announcing our new feature to customers.

You get something like this:

Subject: Exciting News About Our Latest Feature!

Hi there,

We're thrilled to announce the launch of our newest feature! This powerful addition to our platform will help you work more efficiently and achieve better results. We believe this feature represents a major step forward.

We'd love for you to try it out and let us know what you think. If you have any questions, please don't hesitate to reach out.

Best regards,
The Team

You read it and feel that familiar deflation. The grammar is fine. The structure makes sense. But it could apply to any SaaS company announcing anything. It doesn't mention what the feature actually does. It doesn't explain why anyone should care. It doesn't sound like your brand. It has no urgency, no specificity, no personality.

So you sigh, decide AI isn't useful, and write it yourself. Or you search Google for templates and edit heavily anyway.

But here's what actually happened: the AI did exactly what you asked it to do. You asked for "an email announcing our new feature." It gave you an email announcing a feature. It couldn't know which feature, what makes it valuable, who the audience is, what tone you wanted, whether you need urgency or softness, how long it should be, or what action you want readers to take. You didn't tell it any of those things, so it made safe, generic assumptions.

This is the core problem: generic requests produce generic outputs. Not because the AI is stupid. Because you haven't specified what you want. You've just described the category of thing you want—an email, a blog post, a brainstorm list—without the specifics that would let the AI tailor something useful to your actual situation.

The AI is like an extremely literal assistant. An actual assistant would ask you a dozen questions: What period? Which products? What's the main audience—the board, the sales team, investors? Should it be optimistic or brutally honest? How long? What tone? What do you want them to do?

An AI won't ask those questions. It will just make assumptions and hand you what it made.

The Assumption Game

When you give an AI vague instructions, it falls back on patterns in its training data. Thousands of similar requests, thousands of similar outputs from the internet. It settles somewhere in the safe middle. Average. Default. Uncontroversial. Inoffensive to everyone and genuinely useful to no one in particular.

This is why corporate-sounding language dominates AI outputs. This is why so many AI-generated emails start with "thrilled to announce" or "excited to share." This is why brainstorms tend toward the obvious. When the AI doesn't know what you specifically need, it defaults to what works broadly for a general audience—which means it works perfectly for nobody.

Think about it this way: if you asked a speechwriter to write a speech without telling them anything about the occasion, the audience, the speaker, the message, or the desired outcome, you'd get a speech that could fit anywhere and belongs nowhere. They'd have to guess: maybe it's inspirational? Maybe formal? Five minutes long? They'd produce something safe and mediocre because they're working blind.

That's exactly what happens when you ask ChatGPT to "write an email." The AI is working blind.

Who Gets Good Results From AI

People who get good results from AI tend to be the ones who are already good at clearly articulating what they want. Not because they know anything special about AI—they don't. But because they've learned to be specific in other contexts.

A designer briefs a copywriter like this: "We need web copy for an SaaS landing page targeting mid-market operations managers. The tone should feel authoritative but approachable—trusted advisor, not salesman. They're skeptical and busy, so the headline needs to address their biggest pain point: manual processes eating up half their week. We need them to sign up for a demo, not buy directly. Copy should be punchy, max three sentences per section."

That designer gets good outputs because they've provided context, audience specificity, tone, pain point, desired action, and format constraints. When that same designer uses AI, they often get worse results—not because the AI is worse than a professional copywriter, but because they ask the AI less. They think they're saving time by being brief, when actually they're sabotaging the output.

Product managers, consultants, strategists—people used to thinking through complex requests—tend to get better AI results immediately, often without realizing why. They ask questions differently. They specify more. Years of writing briefs or planning documents taught them that clarity is an investment, not a luxury.

Meanwhile, people used to quick, simple requests sometimes struggle with AI. Not because they're less intelligent, but because those previous tools didn't require specification. Google guesses what you want based on context and history. Your phone's assistant knows you're serious about dentist appointments. Those tools hide their assumptions. AI doesn't.

What "Better Prompts" Actually Means

When this book talks about "better prompts," it doesn't mean fancy tricks or secret jargon. It means being more specific about what you actually need. It means thinking through the request the way you'd think through a complex problem at work—by defining the variables that matter.

Take that SaaS feature announcement email again. Here's a better prompt:

Write an email announcing our new workflow automation feature to existing customers. The feature lets them set up custom rules so repetitive tasks run automatically without manual intervention. Our customers are usually operations managers or team leads at mid-market companies—they're busy and skeptical of overpromises. The tone should be confident but not over-the-top. We want them to log in and try the feature within the week. Keep it under 150 words. Subject line should be clear and specific—avoid generic "exciting news" language.

This prompt still isn't perfect, but notice what it includes: what the feature does, who the audience is, what you know about their mindset, the desired tone, the desired action, and a word limit. No vague assumptions. You've pointed the AI toward your specific situation.

Here's what you'd likely get:

Subject: Automate your repetitive tasks—no code needed

Hi,

You just got workflow automation. Set up custom rules in the workflow builder, and tasks like status updates, file moves, and notifications run on their own. No more manual work. It's live for all accounts starting today—check your dashboard to get started.

Questions? Reply here or check out the help docs.

—The Team

This isn't perfect either, but it's specifically useful in ways the first email wasn't. It mentions the feature by name and what it does. It acknowledges the customer segment and their skepticism by being matter-of-fact rather than hype-focused. It gives a clear next step. It's shorter and tighter because you specified a word limit. It sounds like it came from a company that knows its product, not a generic template.

The only thing that changed between these two prompts was your input. The AI is the same. The model is the same. All that changed is that you gave it better information to work with.

The Context Problem

Most people use AI the way they Google. You type what you want, hit enter, see results. If the results are bad, you assume the system failed. What you don't see is how much context you provided without thinking about it.

When you search "best pizza near me," Google uses your location data, your search history, your previous preferences, the time of day, trending restaurants, your profile. Google has massive amounts of context about you and your immediate situation. You barely said anything, but you actually provided huge amounts of information implicitly.

AI works differently. It doesn't have your location history or search patterns or implicit knowledge about your business. It only knows what you tell it, in that specific conversation. Every variable you take for granted has to be made explicit. This feels like extra work because you're used to tools that fill in the blanks for you.

The irony is that this limitation is actually an advantage. Because it forces you to think clearly about what you're asking for. It forces you to become aware of your own assumptions. And because you're being explicit, you can control the output in ways you never could with Google or Alexa.

Why People Think AI is Overhyped

A lot of the "AI is overhyped" sentiment comes from people comparing vague ChatGPT prompts to professional work they've hired people to do. They ask the AI something loose and generic, get back something loose and generic, and then wonder why they can't replace expensive consultants or writers with a chatbot.

Well, they can't—but not because the AI is incapable. They can't because they're not giving the AI the same input they'd give a professional. They'd never call a consultant and say "help us with strategy." They'd schedule a meeting, share financial data, discuss their market and competitors, articulate their constraints and goals. They'd be very specific. Then they'd pay the consultant a lot of money to synthesize all that into a strategic recommendation.

But they ask ChatGPT "give us strategy ideas" in a single sentence and feel disappointed when the results are generic. The comparison is unfair. The AI didn't get the same input the consultant got. So of course the output is worse.

This is actually good news. It means you don't need to hire expensive professionals for everything. But it does mean you need to do the thinking yourself. You need to be specific about your situation, your constraints, your goals, and your audience. You need to articulate the variables. Then the AI becomes genuinely useful—not as a replacement for thinking, but as a tool that helps you execute the thinking you've already done.

The Real Problem Isn't Your AI

If you've been frustrated with AI outputs, the frustration is valid. The outputs have probably been disappointing. But the problem isn't that the technology is broken. The problem is that nobody taught you how to actually use it.

You learned how to Google, and the system learned your habits and filled in the gaps for you. You learned how to use email by typing a message and hitting send. You learned how to use your phone's voice assistant by giving it simple commands: timers, weather, calls, reminders. All of these tools either work in narrow domains where context is obvious, or they learn about you over time and reduce the need for specificity.

AI doesn't work like that. Each conversation is fresh. Each prompt is taken literally. You have to be the one providing context, specificity, and clarity. It's not that the AI is worse than these other tools—it's that they're different kinds of tools, and they require different skills.

The good news: the skill you need isn't technical. You don't need to understand how neural networks work or fine-tuning or tokenization. You just need to learn how to ask a question the way an expert professional would ask it. Specifically. Clearly. With enough detail that the person or system you're asking actually understands what you're looking for.

This is something you already do in your professional life. You just need to do it with AI, and you need to do it more deliberately because the AI won't make helpful assumptions to fill in the gaps.

The Five Elements That Matter

Most advice about AI treats prompting like a trick—like there's some magic wording that will unlock great results. Try starting with "You are an expert in..." or Use the word "think step by step" or Ask it to explain its reasoning. These tips sometimes help a little, but they're treating the symptom, not the disease. The disease is vague thinking. These tricks just dress it up slightly.

A better framework: think about what information an expert would need if you hired them to do this work. If you hired a copywriter to write that email, you wouldn't say "write an email announcing our new feature." You'd tell them what the feature does, who the customer is, what problem it solves, what tone fits your brand, what action you want readers to take. You'd give them examples of your brand voice. You might share a competitor's email so they understand what you don't want. You'd answer questions until they felt they had enough context to do good work.

That's the framework: give the AI what you'd give a professional.

Five specific elements matter consistently across virtually every prompt:

Context: Background information about the situation, the business, the market, whatever is relevant so the AI isn't making wild assumptions.

Specificity: Details about what you want—the audience, the tone, the length, the style, what problem you're solving.

Role: What expertise or perspective you want the AI to bring. A skeptical critic? A creative brainstormer? A direct-response marketer? An operations analyst?

Format: How you want the answer structured. A list? A paragraph? A table? A script? An outline?

Feedback: The willingness to iterate. You'll rarely get a perfect output on the first try, and that's fine—you just refine based on what you got.

Master these five elements, and you can write an effective prompt for almost anything. Ignore them, and even a brilliant AI will give you mediocre outputs because you haven't given it enough to work with.

Why This Matters Right Now

AI tools are now good enough that they're moving from "interesting experiment" to "actual tool people use for actual work." The people who figure out how to use them well will have a huge advantage. Not because they're smarter, but because they've learned how to ask better questions. They'll spend 10 minutes writing a thoughtful prompt and get a solid draft they can refine in 30 minutes, where someone else spends an hour writing something from scratch or asking the AI three times and getting three mediocre answers.

This compounds. Over a month, over a year, over a career, the difference between generic outputs and good outputs becomes substantial. One person gets consistent leverage from AI. The other person keeps getting disappointed and decides AI isn't useful.

The gap isn't talent or intelligence. It's clarity.

That's what this book is about: teaching you to be clear in a way that AI can actually use. Not through secret tricks or magic words, but through thinking like a professional hiring another professional, and communicating your needs accordingly. Once you internalize this framework, you'll stop being disappointed by AI. You'll start being surprised by how good it can be when you actually tell it what you want.