AI Tools to Replace Your $10k+ Creative Agency

This little piece here basically started an entire business. This little prompt formula seemed like this random collection of words with no rhyme or reason — but you’d run it and then you’d get something that made you go, “Oh… okay. I see what’s going on here.”

Hey everyone. Today is the episode that has blown my mind the most this year. We have Rory Flynn on, and he’s showing us how to rebuild the work of an entire design and creative team — photo shoots, all of that — in AI. You can get massive scale. You can create assets incredibly fast for ads, websites, really anything. This is going to blow your mind. Let’s get into today’s show.

Ever wondered how you can use AI to reduce the amount of money you’re spending on ads while acquiring more customers? One of the biggest levers is creative — imagery and video. And we have the perfect person to walk us through that.

We have Rory Flynn, who’s a master at using AI to create vast amounts of creative assets. Rory, welcome to Marketing Against the Grain.

Awesome. Thanks for having me, guys. Really excited about this.

You’ve been grinding on this for a few years now, and you’re going to show us the best of the best — the most important things you’ve learned. I’d love for you to take us behind the curtain. This is where it all started, right?

Yeah. This little thought process right here. If we’re trying to create photorealistic assets — especially on the marketing side — things need to look real. Futuristic spaceships are cool, but we needed stuff we could put in emails.

So we started asking: what actually is a photo? What are the core categories of a photo? What’s controllable inside an image?

This basic structure was how we looked at it. These are what I call the non-negotiables. In AI, if you don’t prompt for them, they’re still going to be filled in. So if you want control, these are the core elements you need to think about.

We’d start with shot type or photo type — perspective, distance, how it’s captured. A close-up versus a drone shot tells a totally different story.

Then subject and action — who’s in it, what are they doing. That’s the core idea of the image.

Then environment — where it takes place. Color scheme — every image has color, even black and white.

Where we really hit early was cameras and lenses. Different cameras create different visual aesthetics. The difference between prompting an iPhone photo versus a Polaroid is massive. Totally different visual signatures.

Then things like film stock. Mood and emotion were always huge for us — the vibe. Lighting is always there whether you like it or not. No lighting means a black image.

So we realized: these things are happening in every image, every time. Why don’t we build structure around them?

We started thinking about them as visual building blocks. If each element is represented in the prompt, iteration becomes fast. If the lighting is wrong, we know exactly where to fix it. If the lens feels off, we know what to change.

That’s huge.

If I’m listening to this, what’s interesting is how much context you’ve built into the prompt. Did this come from photography experience, or did AI help you reverse-engineer it?

A lot of trial and error. I have a graphic design background, so I had maybe half the technical knowledge of a photographer. I didn’t know what focal length was before AI.

But AI forced learning. I’d see an output and think, “I know this look — I just didn’t know the term.” Same with bokeh. I’d seen it my whole life but didn’t know the word.

Your interest grows when you can do real things quickly. Early on, this wasn’t just about business. It was therapeutic. I’d dump my frustrations into Midjourney instead of doom-scrolling Twitter. I was building worlds.

What you’re really saying is: reverse-engineer the thing you want to create. Understand the variables you can control, because those variables determine the output. This thinking works for images, video, music, system prompts — anything.

Exactly. Once you have the building blocks, everything gets easier.

What’s funny is someone watching on YouTube could screenshot that slide, drop it into ChatGPT, and ask: “What are the building blocks of this image?” And in 10 minutes, they’d understand how it works.

100%. Once you have the formula, it becomes simple. The hard part is getting the formula.

We turned those building blocks into a prompt formula — basically Mad Libs for AI. Anyone on the team could fill it in, or we could ask ChatGPT to generate 50 variations.

That little prompt formula basically started an entire business. I didn’t realize how important it was at the time, but I use it constantly now. It’s evolved a million times, but this is the distilled version.

I also learned something important: don’t start with giant five-paragraph prompts. Start condensed. If something’s wrong, it’s easier to troubleshoot one word than three paragraphs.

Start simple. Add complexity later.

Back then, results were slot-machine-like. Now you’re much closer to the result on the first try.

Exactly. For example:
“Motorsport photography, Red Bull F1 car, racetrack, warm tones, 35mm, shallow depth of field, sunset backlighting, center framing, motion blur.”

It sounds like random words. But the output listens. You get the racetrack. You get warm tones. You get blur on the tires. You get shallow depth of field like iPhone portrait mode.

Once you realize it listens, iteration becomes easy.

One question before we move on. That prompt took you a long time to refine. It’s kind of your USP. How do you feel about giving prompts away, knowing someone can shortcut years of learning?

I’m very open about it. I feel like everyone should learn this. I feel a responsibility.

I was a graphic designer who got replaced by Fiverr. I’ve felt this before. A lot of people feel like they might not be needed anymore.

If I can help people understand this, they become valuable again — to companies or themselves. I love getting people to that “Matrix moment” where they suddenly see how it works. Once they get there, they’re fine.

If you want Rory’s prompts and workflows, we’ve put them together for you. You give AI one brief — it spits out 100 assets. Social posts, email copy, blog drafts. You can grab it via the link or QR code below.

Now, you’ve moved on to even more advanced workflows.

Yeah — Weave is my playground now. An image isn’t just an image anymore. It’s the foundation. An image becomes video, 3D, text, code, GIFs. It branches infinitely.

That can be overwhelming, so we narrow it down. One big use case was fast fashion photography.

Brands have thousands of SKUs with short lifespans. Traditional photo shoots are expensive and slow. So we built a workflow to replicate a photo shoot with AI.

This is modular prompting. It looks chaotic, but it’s not.

You create a character — gender, age, body type. You describe the clothing via a system prompt. The model wears the clothes every time.

Then we batch prompts to generate multiple angles: front, side, ¾ view, action poses. We split prompts with a delimiter so they run in parallel.

You’re basically replacing a physical photo shoot with AI.

Exactly. And once you have one image, you branch. One image becomes six angles. Six become sixty.

This is incredibly technical for something traditionally artistic.

Yeah — every field is becoming creative and technical. I love that blend.

This engine lets you change prompts and scale endlessly. You could generate 1,000 images in 20 minutes.

Retail companies should be all over this.

And you’re not even uploading a model photo — the model is AI-generated. Only the clothing is real.

Correct.

Have you done this with video?

Yes — images are precursors to video. I use them as keyframes. That way branding stays consistent across frames.

Text-to-video is great for speed, but image-to-video gives precision. If you control the image, you control the video.

This applies to product design too. Upload your design language and instantly see features across multiple styles.

Exactly. It’s endless. I have to rein myself in.

Another big use case is ad creative. Same batching logic. Generate 10 prompts. Split. Run. Done.

We now add visual brand profiles — dump 20 images into an LLM, extract a style guide, and enforce it via system prompts.

This sounds complex, but it’s just structure.

You’re making thousands of ads in 20 minutes.

Yeah. At that point, the work becomes curation, not creation.

Have you built an agent to rank the outputs yet?

That’s the next problem. AI creates new problems fast.

What people need to understand: you’ve built an entire creative agency workflow with prompts and tools. Agencies charge thousands for three options. You’re generating thousands.

Exactly. It’s custom libraries.

And yes, sometimes I just ask for dumb things like “make a lo-fi meme,” and it delivers because the system prompt is smart.

For B2B, worse images often perform better. We once put a client on a Game of Thrones throne and sold clients for months.

Bad images can win.

The big takeaway: creative teams shouldn’t be gatekeepers anymore. Anyone who needs imagery should be self-sufficient.

Marketing roles have changed. Asset creation, segmentation, testing — everything is 10x now.

So Rory, how much has output increased?

We’ve created more work for ourselves. AI doesn’t remove work — it increases optionality.

Testing is faster. Optimization is faster. Winners emerge faster.

What used to take a year now takes 20 minutes — after three weeks of building the system.

You built a creative team in three weeks.

I’ll take that trade.

The key lesson: start with the outcome. Get one good output. Then reverse-engineer the system to reproduce it at scale.

That’s the only way.

You’re exporting assets and running ads across platforms?

Exactly. I teach teams how to do this. I don’t run ads myself.

This doesn’t replace photographers. It supplements them. It’s for testing, exploration, iteration.

This democratizes creative access. There’s pent-up demand.

It’s an incredible time to be a craftsperson.

Go build.

Rory, thank you so much for being on the show.

Thanks guys. Appreciate it.