Marketing with AI: The Right Way to Use It (and What to Avoid)

9 min readFeb 12, 2025

Marketing involves a lot of things, some exciting, some mind-numbingly repetitive. AI can help, but only if you know how to talk to it.

That’s what I set out to explore in a workshop I recently ran, drawing from my experience leading marketing at Skyflow, my background as an AI researcher, and years of working at the intersection of AI and real-world applications, focusing on where AI actually makes a difference.

I walked attendees through the emerging AI landscape, showed them how to get better results with prompt engineering, and even built a tool — Nimbus Prompter 2000 (because why not give it a Harry Potter inspired name?) — to make the process smoother.

This post is a breakdown of that session: what works, what doesn’t, and how you can get AI to handle everything from ad analysis to writing blog posts without sounding like a robot.

If you’ve ever wrestled with AI-generated nonsense and wondered if it can actually be useful, you’re in the right place. But first, let’s try to answer an important question on many people’s minds.

Is AI Going to Take Our Jobs?

Short answer: no.

Every big technology shift sparks the same fear, machines will replace humans, and we’ll all be out of work. But history tells a different story.

Every major innovation, from the printing press to the Internet, has led to more jobs, not fewer. The work changes, sure, but new opportunities always emerge. AI is no different.

Right now, most AI applications are designed to help humans, not replace them. According to recent analysis from Anthropic, 57% of AI usage is about augmentation, AI assisting workers rather than making them obsolete. That’s the real story. AI isn’t here to do your job for you; it’s here to make you better at it.

Also, as someone who’s been building technology for most of my life, it’s really hard to get any technology to work right 100% of the time. Most of the time, I can’t even get a sensor-activated paper towel dispenser to work in a public bathroom and it’s got one job! So the idea that AI will flawlessly replace complex human roles overnight? Yeah, not happening.

What we should be striving for is Centaur Chess level. This is a type of chess that combines humans with chess playing machines. This concept gained traction after Garry Kasparov, one of the greatest chess players of all time, was defeated by IBM’s Deep Blue.

Instead of human versus machine, Centaur Chess reimagines the game as human plus machine. In these matches, human players team up with chess engines, combining intuition, strategy, and creativity with AI’s ability to calculate millions of possible moves. As a result, even an average player with an average chess engine can beat the most powerful supercomputer running solo.

Why?

Because human intuition combined with AI’s computational power is stronger than either one alone. That’s the model we should be aiming for in marketing (and everything else, really).

So, AI won’t take your job, but marketers who know how to use AI might.

The Marketing Tech Landscape is Changing — Fast

AI-first marketing stacks are taking over.

In the next few years, tools like albert, Surfer SEO, Sprouts, and Regie will be everywhere, helping marketers with ad optimization, analyzing data, and optimizing content faster than ever. The landscape is shifting, and AI-driven tools will become the default.

That said, you don’t need a stack full of AI-powered software to start using AI effectively. With good prompt engineering, you can get a lot done on your own. Writing ad copy, improving messaging, brainstorming content, if you know how to structure a prompt, you can turn AI into a useful assistant without relying on specialized tools.

So before getting overwhelmed by the latest AI-powered marketing software, let’s start with the basics: how to get AI to actually do what you need, just by asking the right way.

How to Get AI to Actually Do What You Want: A Guide to Prompt Engineering

To get the most out of AI, you need to know how to ask the right questions. This is where prompt engineering comes in, the skill of crafting instructions that guide AI toward useful, relevant responses.

Do it well, and AI can be a powerful tool. Do it poorly, and you’ll get vague, generic answers that don’t really help.

But before jumping into prompts, it helps to understand a little about the strengths and weaknesses of LLMs.

A Bit About LLMs

AI models are great at some things and terrible at others. Knowing these strengths and weaknesses makes a big difference in how you prompt them.

They lack real-world experience. LLMs don’t “know” things the way people do. They don’t have up-to-date knowledge or personal experience. Ask about last night’s Lakers game or a feature your company plans to launch, and you’ll get nothing useful, unless you give it that information.
They’re great at certain tasks. LLMs excel at pattern recognition, language generation, and summarization. They can rephrase things in different styles, analyze sentiment, and generate text quickly, but they’re not fact-checkers or original thinkers.
They need direction. AI can be creative, but without clear guidance, it can also go off the rails. It doesn’t automatically know what’s relevant to your goals, you have to steer it.

The Art of Prompting

If you give AI vague instructions, you’ll get vague results.

🚫 Bad prompt: “Write something catchy for our new product.”

You’ll likely get a generic response. It’s like asking a new intern on day two to write something catchy. They’re afraid of getting fired so they’ll say yes, but they have no idea what you actually want.

✅ Better prompt: “You are a copywriter for a tech company launching a new smartwatch. Write a catchy Twitter post announcing the product.”

This is more specific, but catchy is still subjective. What kind of catchy? Funny? Bold? Minimalist?

When you’re just starting, the best approach is to iterate. If you get something decent, refine your request. Give examples. If you like a certain style or phrasing, ask for more like it. If you hate a response, tell AI what to avoid.

Frameworks for Better Prompts

There are a lot of ways to structure a good prompt, but one useful framework is COSTAR:

Context — What’s the background information AI needs?
Objective — What do you want to achieve?
Style — Should it be formal, casual, humorous?
Tone — Playful, authoritative, professional?
Audience — Who is this meant for? Who am I targeting?
Response — What format do you want? A list? A paragraph? A headline? Twitter or a blog post?

You don’t need to include all of these every time, but thinking through them helps make AI’s output more useful.

Next, let’s go through some real examples of prompt engineering in action.

Pre-Built Prompts for Marketing

To make prompt engineering for marketing easier, I built a tool that has all these prompts pre-loaded, along with different personalities to shape the style and tone of responses.

If you want to see it in action, check it out here: Nimbus Prompter 2000.

Nimbus Prompter 2000 — Prompt Engineering for Marketers

This tool includes 12 marketing-specific prompts that help structure AI-generated content for different use cases, everything from ad campaigns to SEO analysis.

Let’s break down one of them to see how it works.

Breaking Down a Marketing Prompt: Ad Campaign Example

One of the prompts I created helps generate ad copy for a new product launch. It follows the COSTAR framework, ensuring we are setting the right context, objective, and structure for a useful response.

Ad Campaign Prompt

Context:
You are the lead copywriter for KubeEase, a revolutionary iPaaS platform that
automates the deployment and scaling of Kubernetes applications.

Our brand voice is innovative, approachable, and empowering.

We're launching a new feature: AutoPilot, which simplifies Kubernetes scaling
with intelligent auto-scaling and zero-downtime updates.

Target Audience:
Developers, DevOps professionals, and tech startups aged 25–45 who want the
power of Kubernetes without the complexity. They value innovation, time-saving
solutions, and reliability in their workflows.

Campaign Goal:
Increase awareness of the new AutoPilot feature and drive signups through
targeted social media advertising.

Competitive Landscape:
Our main competitors are CloudFlex and K8sNow. CloudFlex offers a broad suite
of cloud tools but is costly and complex. K8sNow is affordable but lacks
advanced auto-scaling and user-friendly features.

Instruction:
Create a series of three short, compelling ad copies (max 50 words each) for
LinkedIn that highlight the simplicity and power of AutoPilot for Kubernetes.
Each ad should include a clear call-to-action.

Input:
Product USPs:
Automates Kubernetes scaling with intelligent load balancing.
Enables zero-downtime updates with smart rollbacks.
Simplifies Kubernetes management for teams of all sizes.
Flexible pricing with no hidden fees.

This structured approach ensures the LLM has everything it needs to generate relevant, high-quality ad copy.

With Great Power Comes Great Responsibility: The Dos and Don’ts of Prompting

AI can be a powerful tool, but like any tool, it’s only as good as how you use it. Getting great results from AI isn’t just about writing good prompts, it’s about knowing how to refine, fact-check, and personalize what comes out.

Here’s how to use AI effectively without falling into common traps.

✅ Do This

Review and refine outputs. As is the case with your own writing, AI-generated text isn’t perfect on the first try. Tweak your prompt, experiment with different instructions, and refine the results to get something actually useful.
Personalize the output. AI can get you 80% of the way there, but that last 20%, your voice, insights, and expertise, makes all the difference. Add your own touch.
Fact-check everything. LLMs are like a toddler that’s read everything on the Internet, it knows a lot of stuff, but not always a reliable storyteller. They’ll confidently make things up, so never take their output at face value.
Use AI as a non-judgmental learning tool. Need a complex topic broken down? Ask AI to explain it like you’re five. Ask basic questions without fear of looking dumb. AI won’t judge you.

❌ Don’t Do This

Don’t share private information. Once you put something into an AI tool, you can’t take it back. Be mindful of what you share.
Don’t blindly trust AI’s output. It bears repeating, LLMs get things wrong. Fact-check before you use anything.
Don’t copy and paste AI-generated text without edits. AI should assist, not replace. Use it to speed up your work, not to take over completely.
Don’t “delve.” Some AI tools (especially ChatGPT) love the word delve, and certain other overused phrases. If your text sounds robotic or like a generic essay, rewrite it.
Don’t start with “In this digital landscape.” AI has patterns. If something sounds cliché or unnatural, it’s probably because AI has written it that way a thousand times before. Break the pattern.

Good AI use is about collaboration, not automation. The best results come when you treat AI like an assistant, not a replacement. Keep it honest, keep it human, and make it work for you.

Wrapping Up: AI is a Tool, Not a Replacement

AI can’t do your job for you, but it can make you faster, sharper, and more effective — if you know how to use it. The key is good prompting, smart refinement, and knowing the strengths and limitations of the tools.

Whether you’re crafting ad campaigns, analyzing leads, or brainstorming content, AI can be a powerful assistant. Just remember: review, personalize, fact-check, and don’t sound like a robot. And if you want to skip the trial-and-error, check out Nimbus Prompter 2000 to get started with pre-built marketing prompts.

Now go make AI work for you, not the other way around.

Note: If you’re interested in the code for Nimbus Prompter 2000, you can check it out here.