Get the AI Tipsy for 10X Better Results

The Concept: The AI Performs Better After a Few Shots

Let me share a story about a veteran content manager with 7 years of SEO experience. We’ll call her Lisa.

Lisa just hired a junior writer. This writer is eager, smart, and capable. But she’s completely new to the nuances of writing meta descriptions (those little blurbs you see under titles in Google search results) that actually drive clicks.

On day one, if Lisa asks the junior to write meta descriptions for their company’s blog posts, the junior will likely create something technically correct but totally forgettable. It’ll be the SEO equivalent of elevator music: safe, generic, and invisible in search results.

But what if Lisa sits down with this junior and shows them three of her best meta descriptions? She points out how she always starts with an action verb, naturally weaves in the keyword without stuffing, creates urgency without being pushy, and stays within the 158-character limit while making every word count.

Suddenly, that junior writer gets it. They see the pattern. They understand exactly what Lisa is looking for.

This is exactly how AI works.

By default, it gives you the “first day junior writer” version: cautious, generic, technically correct but missing that expert touch you’ve developed over years. This is called a Zero-Shot Prompting. You’re asking it to perform cold, with zero examples to guide it.

But when you give it a few “shots” of examples first, showing it exactly what excellence looks like, you transform that generic AI into a trained assistant who can replicate your hard-won expertise at scale. This Few-Shot Prompting technique is how you teach AI to perform like someone you’ve personally mentored, not like a forgettable generalist.

Let me show you how this technique works.

How It Works: Show Me, Don’t Tell Me

The principle of few-shot prompting is all about showing the AI what you want instead of telling it.

Most people start out with a zero-shot approach (like the newly hired junior assistant):

Write a meta description for a blog post about content marketing ROI.

You can’t get more generic than this prompt.

I could have easily given the AI some rules and instructions to follow for a meta description. But for the sake of this example, I’m not providing any guidance.

I just want you to see what happens. Here’s the output:

Learn how to measure content marketing ROI with proven metrics and formulas. Discover strategies to track performance, justify budgets, and boost returns.

Wow.

That’s got to be the lamest meta description I’ve ever read.

It’s more boring than two plain pieces of bread slapped together and called a sandwich.

Okay, so if that was zero-shot prompting (no examples), let’s look at what few-shot prompting looks like:

Write a meta description for a blog post about content marketing ROI.

Here are some examples of meta descriptions that I like:

—
Blog Post: B2B Email Marketing Best Practices
Meta Description: Transform your cold emails into qualified meetings with proven B2B email tactics that the top 1% of use to achieve 3x higher response rates. Learn what actually works. 🎯
—
Blog Post: Content Calendar Template
Meta Description: Steal the exact content calendar template that helped 5,000+ brands triple their engagement in 30 days. Pre-filled formulas, automation hacks included. 📅
—
Blog Post: SEO Keyword Research Guide
Meta Description: Unlock the exact keyword research framework that drives 73% of organic traffic to top-ranking sites. Master profit-focused keyword selection. 🔍
How to Find Your Dream Customers
—
Blog Post: How to Find Your Dream Customers
Meta Description: Stop wasting money on wrong-fit clients. Learn the exact framework $10 million dollar agencies use to magnetically attract customers who pay premium prices. 💎
—
Blog Post: 10 Types of Cold Emails That Convert
Meta Description: Master the cold email templates top sales reps guard like trade secrets—from pattern interrupts to the value-stack closes that turns 50% of ice-cold prospects into eager buyers. 🔥
—

So it’s the same prompt, but I’ve given the AI some examples of meta-descriptions that I like (along with the corresponding blog post title for each).

It’s easy to make meta-descriptions, so I gave 5 examples to the AI. My rule of thumb with few-shot prompting is at least 3 examples, but given how powerful modern AI models are, you don’t need more than 7 to 10. Once you go past that, the returns are negligible and can actually start harming your output if you’re not careful.

Now here’s what I did with my examples: I added some numbers to each along with an emoji at the end (yep, those are allowed by Google in meta-descriptions and help grab eyeballs).

Also look how I separated each example with a triple “—“. It’s important to use some kind of delimiter to separate the examples, otherwise the AI can get confused.

So let’s see if the AI can pick up on the numbers and emoji without me having to explicitly tell it:

Steal the ROI calculation framework that helped 500+ brands prove content marketing drives 6x more revenue than paid ads. Turn content skeptics into believers with hard data. 📊

Look at that!

It sounds way better than the zero-shot result, and it even has numbers and emoji at the end!

That’s fantastic!

This was honestly a super simple example. I probably wouldn’t use few-shot prompting to create a meta description; I’d just follow some basic rules about what I wanted. But I wanted to show you how this works with something really basic.

How about we try something a little more interesting?

Something a Little More Interesting

Few-shot prompting really shines when you can’t easily explain what you want using just words.

Sometimes you want the AI to match a specific writing style or tone, or maybe there’s another concept that’s tough to describe. In those cases, it’s much easier to show the AI some examples than try to explain it with words.

Let me give you an example of what I mean with this prompt:

Write a thought-provoking essay about “no one truly understands me”.

The Unbridgeable Distance: On Being Fundamentally Unknowable

The sentiment “no one truly understands me” resonates across generations and cultures with such consistency that it might be considered a defining feature of human consciousness. It emerges in teenage diaries and deathbed confessions alike, whispered in therapy sessions and shouted in arguments. Yet for all its universality, this feeling points to one of the most profound paradoxes of human existence: we are simultaneously desperate to be known and convinced of our essential unknowability.

At its core, this belief rests on an undeniable truth about subjective experience. The texture of your thoughts, the particular shade of your melancholy on a Tuesday afternoon, the way certain memories taste in your mind—these inner phenomena are yours alone. No amount of language, no feat of empathy, can fully transport another person into the private theater of your consciousness. We are each locked in what philosophers call the “phenomenological bubble” of our own experience, pressing our hands against the glass, seeing others but never quite touching.

This isolation becomes more acute when we consider how much of our inner life remains opaque even to ourselves. We carry within us a vast unconscious realm of drives, fears, and associations that influence our behavior in ways we barely comprehend. If we cannot fully understand ourselves, how can we expect others to understand us? The complaint that no one truly knows us might sometimes be a projection of our own self-mystification, a displacement of the unsettling fact that we are strangers to ourselves.
Yet the universality of this lament contains its own strange comfort. If everyone feels misunderstood, then perhaps this very feeling is something we share, a common thread in the human tapestry.

….

I stopped it at “tapestry”. Man, I hate that word. It’s such a nasty AI word from 2023.

In my opinion, that result is hot garbage. Who would even read it? It’s like the AI is trying to use “big” words to sound intelligent.

So, what am I trying to achieve? Let me show you (and the AI) with some examples:

Write a thought-provoking essay about “surrendering”:
Surrender and Eat Pancakes

There are four things you need to make good pancakes:

Flour, Butter, Milk, Eggs

You can find these ingredients anywhere. But there is only one place where you’ll find the right combination of them: in the kitchen of your mom or grandma.

Mom used to make pancakes for breakfast, lunch and dinner. She made them from scratch, using the best flour she could afford, butter that was fresh from the dairy case, milk straight from the cow, eggs that were laid by chickens who ate grass and corn and bugs and bugs and bugs.

And she made them with love.

I’m sure you can imagine that this recipe is not written in stone. There are variations. Some days, she might use whole wheat flour instead of white, or add raisins or chocolate chips or even peanut butter.

And sometimes, she’d make them with frozen blueberries or strawberries instead of raspberries.

This is true of recipes, life and business.

Our lives are filled with choices. Sometimes the choice seems obvious, other times it feels like we’re being forced into a corner. We can either choose to fight, or we can choose to surrender.

Surrendering is easy when it comes to pancakes. It’s hard when it comes to a career. But it’s the path to success.

So let’s stop fighting. Let’s start surrendering.

Let’s eat some pancakes.
——————————————
Write a thought-provoking essay about “rebirth”:
The Power of New Beginnings

Every day, every week, every year, we face a new beginning. A new opportunity. A new challenge. A new chance at a different life. But, as much as we may wish it, none of us starts over.

We carry around the past, and we live in the present. We are always walking in our own footsteps, carrying the same baggage. We have no idea where we’ve been, only where we’re going.

Some of us have had the experience of starting over. Others haven’t. In fact, the vast majority of us have never experienced anything remotely close to a new beginning.

Why?

Because we don’t see ourselves as capable of doing so. We don’t think we deserve it. Or, perhaps, we just assume that the new beginning will be worse than the old. Either way, we don’t believe we deserve to move forward, and we certainly don’t believe we have the capacity to move forward.

That’s crazy.

Think about it: If you don’t believe you deserve better, then you won’t try harder. If you don’t believe that you have the ability to create better, then you won’t try harder.

If you don’t believe you can change, then you won‘t try harder. We all want to do more. We all want to be more. We all want more out of life.

It’s time to change our minds. It’s time to give ourselves permission to begin again. To start over.

And it’s time to embrace the power within us to do just that.

When someone asks you why you aren’t more successful, how do you respond? Do you argue that you don’t deserve any better? Or do you explain that because you lack the necessary skills and resources, you simply cannot achieve more? Or maybe you blame circumstances beyond your control.

None of those excuses sound very convincing, do they?

What if I told you that you already have everything you need to succeed? You already have all the tools you need to become the person you want to be. All you need is the courage to step outside of your comfort zone and take action.

That’s what I call The Power of New Beginnings.

The Power of New Beginnings is the belief that you have the capacity to change. That you can create a new reality. That you have the ability to make yourself into exactly the kind of person you want to be, regardless of your current situation.

I know you have it inside of you. I know you have everything you need to rise above your circumstances. All you need to do is take that first step toward a new beginning.

So what are you waiting for? Are you ready to start over? No matter where you are now, no matter what your story is, no matter what you think you should be doing, you can change.

You have the power to reinvent yourself.

Start today.
——————————————
Write a thought-provoking essay about “success”:
A Success Mindset

Success is never final. Success is never permanent. Succeeding is about constant improvement, so nothing stands still.

Nothing goes unchanged. Every time we achieve victory, a new challenge presents itself. When we fall short, a new learning takes shape.

Successful people see setbacks and failures as opportunities to learn. They recognize failure as temporary, they adapt their strategy and they keep moving forward.

Successful people believe that if you fail, you’ve learned more than if you had succeeded.

This mindset is required for any successful endeavor. Whether you are taking baby steps towards your dream or making real progress, you must see each accomplishment as just another milestone on your journey.

This mindset will help you move beyond obstacles and give you the energy to overcome adversity.

When you adopt a “success mindset” you won’t feel bad about failing, because you expect to fail. Failure is just part of succeeding.
——————————————
Write a thought-provoking essay about “the beauty of life”:
Life Is Fragile

We live in a world of uncertainty. Our lives are unpredictable. Our relationships are fleeting. Our health is precarious. Our jobs are uncertain. Our futures are unknown.

This isn’t a statement of pessimism. This is a statement of realism.

Our lives are fragile. Life is fragile.

No one knows when they’ll die. No one knows when they’ll lose a loved one.

No one knows when their job will end. No one knows when the next heart attack will strike.

But fragility is not weakness. It’s strength.

It means we’re resilient. It means we can recover from mistakes. It means we can bounce back. It means we can survive anything.

Fragility is the greatest gift we can ever receive.

Fragility is why we can never stop trying. Because no matter how hard things get, no matter how much we struggle, no matter how far we fall, we can always get up again.

And that’s the beauty of life.
——————————————
Write a thought-provoking essay about “I feel sad”:

Before you look at that result below, how would you describe the writing style I’m going for?

It’s pretty tricky to explain, right?

I actually asked AI to describe it and got this response:

“It uses simple, conversational prose mixing everyday metaphors with life philosophy. Short, punchy sentences create rhythm. Direct, motivational tone using ‘you’ to engage readers. Unpretentious language building to inspirational calls-to-action.”

But even when I gave the AI those exact instructions:

“Write a thought-provoking essay about ‘I feel sad’ using simple, conversational prose mixing everyday metaphors with life philosophy. Short, punchy sentences create rhythm. Direct, motivational tone using ‘you’ to engage readers. Unpretentious language building to inspirational calls-to-action.”

It still couldn’t quite write something close to those examples.

So the only way for the AI to write in that style, is to use those examples as few-shot prompts.

Here’s the output from that long prompt above:

I feel sad.

Three words. Simple. Honest. Human.

We’ve all said them. We’ve all felt them. Sometimes whispered into a pillow at night. Sometimes screamed into the void. Sometimes typed into a text message we never send.

Sadness is heavy. It sits on your chest like a wet blanket. It makes your bones ache. It turns food into cardboard and colors into gray.

But here’s what no one tells you about sadness: it’s supposed to be heavy.

Think about it. When you hold something light, you barely notice it’s there. A feather. A piece of paper. A bubble. They float away, forgotten.

But when you hold something heavy – a stone, a book, a sleeping child – you feel its weight. You know it’s real. You can’t ignore it.

Sadness is heavy because it matters.

It’s heavy because you’ve lost something worth mourning. Because you’ve hoped for something worth wanting. Because you’ve loved something worth missing.

The weight of sadness is proportional to the weight of what it represents.

When you say “I feel sad,” you’re not admitting weakness. You’re acknowledging that you’re brave enough to care about things. That you’re human enough to feel loss. That you’re alive enough to know the difference between what is and what could have been.

Some days, the sadness feels unbearable. Like you’re carrying the entire ocean in your chest. Like gravity has doubled just for you.

On those days, remember this: every heavy thing you’ve ever carried, you’ve eventually put down. Not because it disappeared. Not because it became lighter. But because your arms got stronger. Because you learned where to place your hands. Because you figured out how to shift the weight.

Sadness works the same way.

You don’t have to pretend it’s light. You don’t have to smile through it. You don’t have to rush through it.

You just have to carry it until you’re ready to set it down.

And when you do – when you finally feel that weight lift – you’ll realize something remarkable: the sadness didn’t make you weaker.

It made you someone who knows how to carry heavy things.

Someone who understands that feelings have weight because they have worth.

Someone who can look at another person and say, “I know that weight. Let me help you carry it for a while.”

I feel sad.

Three words. Simple. Honest. Beautiful.

Because they mean you’re still here. Still feeling. Still human.

And that’s everything.

I’m at a loss for words. The output is incredible! It actually touches my heartstrings a bit.

What do you think?

So yeah, that is the power of few-shot prompting and getting the AI tipsy.

Caveats to Few Shot Prompting

Now at this point, I’d usually include some advantages or tips about when to use few-shot prompting. But you’ve probably figured that out by now.

So let me share some warnings instead. Few-shot prompting is one of the trickier principles that you can mess up without understanding why.

Trust me, I’ve been there.

Back in 2023, I spent about 600 hours few-shot prompting an AI model called GPT-J. It was so dumb that few-shot prompting was the only way to get decent results from it.

I learned a lot from that experience, so I want to give you some tips before you embark on a drunk adventure with the AI.

Good Output Requires Good Examples

This is probably the biggest problem with few-shot prompting. You need good examples that the AI can learn from if you want high-quality output.

Some people try to cheat by asking the AI to create the examples. But that doesn’t work well because, remember, we’re trying to get the AI to generate content it can’t make with just text commands.

To get truly good examples, you’ll have to write them yourself or hire someone to write them for you. You might get lucky and find some old human-written examples lying around your office.

I know it sucks, but this part is mandatory. There’s no way around it.

Every Word in Your Example Matters

Every word matters when you’re few-shot prompting the AI.

Say that three times with me:

Every.

Word.

Matters.

The AI looks for patterns. It tries to find them everywhere.

Why is this word here? Why does this sentence have an exclamation point? What’s this grammatical mistake doing here? Ooh, an emoji! I love emojis! What about word count in sentences? Yep. And number of sentences in a paragraph? Yes, that’s part of it too.

The AI is like a dog that hasn’t been out of the house in a year. You let it loose, and it’s out there sniffing everything, including other dogs’ butts.

So a lot of this comes down to having good examples.

What advice can I give you? Start being very analytical and detail-oriented with your examples. Then edit your examples, edit, and then edit some more.

Add One Example at a Time

You might be tempted to add 10 examples, see what happens, and then edit all of them at once because the AI’s output sucks.

Don’t do that.

That’s like taking 50 supplements you’ve never tried before, all at once, and then getting sick.

Which supplement made you sick? You won’t know because you did something stupid. You should take one supplement at a time, give it a week, and write down any benefits or problems you notice.

The same goes for few-shot prompting.

Start with one example. Get it working as well as you can. Do lots of tests. Check what you get back. Once you’ve squeezed everything you can from it, add another example. Then repeat the process.

This is exactly how I approach few-shot prompting. It’s not fun, but it’s the only way to get the highest-quality output that destroys what generic prompts give you.