We all know that LLMs are trained on general best practices. The internet's collective idea of "what good looks like." So when someone on your demand gen team opens ChatGPT and asks it to write a LinkedIn post, or generate a slide, or produce a one-pager (without any brand context, guardrails, or specific rules) the model does exactly what it was built to do. It gives them something that looks and sounds competent enough but that is, by definition, average.

That output goes into the world. It (hopefully) gets engagement. It gets scraped. It reinforces the model. The next person gets something slightly more mediocre. And the one after that. And the one after that.

Now they're not just producing mediocre content anymore. They're actively training the next generation of mediocre content to be even more mediocre. It's AI inception (I'll take my check now, Chris Nolan) and your competitors are doing the exact same thing. So... congratulations, you all look the same now!

How does one go about trademarking “AI inception”? Asking for a friend…

I've been in a lot of conversations with enterprise marketing leaders lately about how their teams are actually using AI day to day. The same thing keeps coming up. There's pressure from "leadership" to use more AI so teams dive head first into experimenting with Claude, ChatGPT, Gemini, Cursor, and so on. The work seems passable at first. Then more people start using the tools and the non-designers get their hands on it and eventually the content looks nothing like the brand. It's not because the tools are bad—they're doing their job with the information available to them. It doesn't look like the brand because there's nothing there to tell Claude what the brand actually is.

That's not an AI problem. That's a context problem.

What is a context layer?

Depending on your perspective, the good (or bad) news is that the technology isn't going back in the box and it's only going to get better, and become more deeply embedded into how teams work. So the question isn't whether to use it. The question is have you given AI what it needs to be successful?

Spoiler alert: most of you haven't and that's ok.

Many of you have seen random markdown files floating around on LinkedIn that seem cool because they have weird symbols and hex values and look like code. But there are two levels of context setting, and almost everyone stops here at the first level.

Level one is what most teams call "setting up our brand for AI." Colors. Typography. Logo files. Maybe a tone of voice paragraph. This takes about twenty minutes, feels productive, and is about as effective as that brand guidelines PDF collecting dust. The AI now knows your primary color is blue. So you get blue colored best practices in your font of choice. None of this actually helps your brand.

Level two is where the magic happens. It's the stuff that takes time and expertise and looks boring from the outside (tbf it looks boring from the inside too) but is the entire reason it works:

• Color rules that go beyond palette. Not "these are our colors" but "Here's how to use our colors. These colors go together. These don't. Here's why and here's when."
• Icons and photography that are indexed, tagged, organized, and described so AI can pick and choose them based on goals instead of vibes.
• Character counts, sizing rules, and design behavior baked into the brief so what comes out actually fits where it's going.
• Canva templates with data autofill tags mapped to specific config criteria, so the system makes decisions humans don't have to.
• Project and skill files built directly into the tools your team uses every day, so the context travels with the work.
• Connectors, MCPs, and webhooks so that when someone does their thing, the brand just works.

This is the context layer. This is the brain. Without it, AI falls back on best practices and produces content that is indistinguishable from your nearest competitor's indistinguishable content. With it, AI runs on your specific guardrails, your specific rules, your specific brand. People stop having to wonder about whether something is on-brand. The system takes care of that for them so they can fully focus on the task at hand.

Or you know... idiot proofs it for humans and machines.

You get what you pay for

The teams getting this right aren't the ones who adopted AI fastest. They're the ones who built the system first and let AI run on top of it. Some fell into it because they already had a robust design system to begin with. Some understood that the work coming out wasn't good enough and knew there was a better way.

The teams getting it wrong saw AI as a replacement for the people whose job it was to build and scale the system. They started making cuts and told everyone to "just use AI for that." And now they have more content than ever while blending further and further into the middle. This is how household brand names become invisible. It comes down to the making the wrong bets.

The bets themselves aren't about whether to choose Claude or Gemini—those are just tools. The bets come down to whether you want to invest in setting the right context or whether you trust that the AI is good enough on it's own. I'm not betting on the latter. Trusting in AI alone is your fastest path to the middle of the pack.

A rising tide... and all that

So rather than compounding mediocrity, we can set the right context to compound quality instead. By building the right systems we're not just protecting our brands, we're forcing everyone to step their game up.

A year ago AI was a nice to have for most businesses. Today it's the first question we get asked on every call. AI is here to stay. The tools are going to continue learning. The question is will they learn what you want them to or what everyone else does.

If you don't know where to start, the good news is OhSNAP! does.