Show, Don’t Just Tell

Using examples to guide AI

A quick story: the contractor problem

Picture this: you’re explaining what you want to a contractor.

“Something modern. Clean lines. Contemporary but warm.”

They nod politely. But you can feel it happening—they’re translating your words into their version of modern. Their taste. Their assumptions. Their mental Pinterest board.

Then you pull out a photo and say, “Like this.”

And suddenly everything changes. In one image, you’ve communicated tone, materials, vibe, and priorities. Ten minutes of description collapses into a shared understanding.

That’s the power of examples. And it’s the same power that makes AI dramatically easier to work with.

Why words alone make AI guess

When you prompt an AI with words like:

  • “professional”
  • “friendly”
  • “modern”
  • “high-end”
  • “casual”

…you’re not giving instructions so much as you’re giving labels. Labels are slippery. “Professional” to one person means formal and buttoned-up; to another it means clear, confident, and simple. “Casual” could mean playful and emoji-filled, or it could mean just less stiff.

AI models are trained to predict what usually follows those words across millions of contexts. So if your prompt is mostly abstract, the AI has to pick a likely interpretation.

That’s not a flaw—it’s literally the job. But it means vague prompts create “competent averages,” not your specific taste.

Examples skip the interpretation step.

When you show the AI something you like, it doesn’t have to guess what “modern” means to you. It can observe what you mean by modern: sentence length, rhythm, structure, word choice, formatting, even how direct (or subtle) the voice feels.

In short: examples compress a lot of meaning into a small package.

The before-and-after difference

Here’s what many people do first:

“Write me a product description. Make it engaging and modern.”

You’ll probably get something usable. It will sound like a product description. It may even be lively.

But then comes the real question: is it your kind of engaging?

Now compare it to this approach:

“Here’s a product description I love. Match that style—short punchy sentences, clear benefits up top, and speak directly to the reader. Keep the structure, but rewrite for my product.”

This version gives the AI a target, not just a direction.

A direction says, “Go north.” A target says, “Go to that building right there.”

If you want fewer rewrites, fewer “almost but not quite” drafts, and less frustration, examples are the fastest way there.

The Example Sandwich (a simple method that works)

A helpful way to think about this is cooking.

A recipe might say: “Sauté until golden.”

But if you’ve ever cooked, you know “golden” can mean five different shades depending on who’s teaching. Watching someone cook shows you what “golden” actually looks like in practice.

Examples do the same thing for tone and style.

Try this structure, which I like to call the Example Sandwich:

  1. State what you need
  2. Show an example
  3. Specify what to adapt (and what to avoid)

Here’s what it can look like in real life:

  • Need: “Write an email announcing a schedule change.”
  • Example: “Here’s an email with the tone I like.”
  • Adapt: “Keep it warm and direct, but shorten it by 30% and remove jokes.”

The “sandwich” works because it combines clarity (your instructions) with reality (the example).

What makes a good example (and how to pick one fast)

Not all examples help equally. The best ones do two things:

1) They match your domain

If you’re writing a landing page, show a landing page.
If you’re writing a personal bio, show a bio.
If you want LinkedIn-style thought leadership, show that—not a casual blog post.

The closer the example is to the real task, the less translating the AI has to do.

2) They highlight what matters

When you say “Write like this,” you might mean:

  • the tone (warm, grounded, bold)
  • the structure (hook → story → bullet points → CTA)
  • the length (tight and skimmable)
  • the reading level (simple, human, jargon-free)

Spell out which parts you want copied. Otherwise, the AI may imitate the wrong thing—like duplicating formatting when you mainly cared about the voice.

Bonus: show what you don’t want

Contrast is powerful. If you can include a quick “not this” example, you refine the target even more.

For instance:

  • “Don’t sound like corporate PR.”
  • “Avoid hypey marketing language.”
  • “No длинные paragraphs or overly academic tone.”

Even a single sentence of what to avoid can prevent a whole draft from drifting.

The real payoff: fewer revisions, more “yes, that’s it”

That contractor could have built something from your adjectives alone—and you might have spent weeks correcting the results.

The photo didn’t replace words. It supported them. It carried information words couldn’t easily carry.

AI works the same way.

When you combine clear instructions with a strong example, you stop “prompting” in the vague, wishful sense and start directing in a practical one. You’re no longer hoping the AI reads your mind—you’re giving it something concrete to align with.

Closing thought

Words are useful. But examples are efficient.

So the next time you catch yourself typing “make it sound more modern” or “more professional,” pause and ask: Can I show what I mean?

Sometimes the fastest way to communicate is still the simplest:

“Like that. That’s what I’m after.”

Discussion