What happens when AI starts using your brand guidelines?

Brand guidelines have always been written for people. But what happens when the people creating content increasingly rely on AI?

Published Read7 min read CategoryChange
A person stands among illuminated gold lines forming a vast digital network against a dark background

Brand guidelines have always been written for people. But what happens when the people creating content increasingly rely on AI?

It’s something I’ve been thinking about quite a bit over the past year.

We’ve spent decades creating brand guidelines. Beautifully designed PDFs, carefully documented typography, colour palettes, photography styles, tone of voice and countless examples of what you should and shouldn’t do.

And for the most part, they work.

The problem is that they rely on someone actually reading them, understanding them and applying them correctly.

Anyone who’s worked with a large organisation will know how inconsistent that can become.

One team interprets the guidelines slightly differently from another. Someone creates a presentation using an old template. Marketing produces something that looks great but doesn’t quite follow the brand.

It’s not necessarily because people aren’t trying. It’s because maintaining consistency across an organisation is genuinely difficult.

Now introduce AI into that equation.

AI doesn’t automatically understand your brand

Over the past couple of years, we’ve seen AI become increasingly capable of producing content, designing interfaces, building websites and generating marketing materials.

Tools such as Claude, ChatGPT and Codex can produce some impressive results.

But there’s a fairly fundamental problem.

They don’t automatically know your brand.

You can upload a brand guidelines PDF, provide a few instructions and hope for the best. Sometimes you’ll get something reasonably close.

Other times, you’ll get something that looks nothing like the brand you’re trying to represent.

And the more people across an organisation use different AI tools, the greater the potential for inconsistency.

It made me wonder whether we’ve been approaching brand guidelines in the wrong way.

What if brand guidelines were written for AI as well?

Rather than treating brand guidelines as a document that someone needs to interpret, what if we made them something AI could actively work with?

Not just a collection of logos and colour values.

A structured system of instructions that explains how a brand should behave.

How typography should be applied. Which colours are appropriate in different contexts. How imagery should be selected. What the tone of voice should sound like.

And importantly, what shouldn’t happen.

I’ve been exploring this approach through something I’ve been calling the Brand AI Toolkit.

The principle is relatively straightforward.

Take the information traditionally contained within brand guidelines and translate it into structured, machine-readable instructions.

Markdown files form part of that foundation, alongside design tokens, reusable components, templates and supporting assets.

The idea isn’t to replace the original guidelines.

It’s to make them considerably more useful.

Taking the idea into a real business

More recently, I’ve been applying this thinking while working with Defaqto.

Alongside developing the foundations of its digital brand system, I’ve been exploring how those same foundations can be interpreted and used by AI tools.

This has been particularly interesting because it’s moved the concept beyond a theoretical exercise.

It’s one thing to demonstrate AI producing something that vaguely resembles a brand.

It’s another to give it the structure and instructions needed to produce something that can realistically be used by a business.

There’s a considerable difference between the two.

And it’s also highlighted that the quality of the underlying brand system becomes even more important.

If your guidelines are inconsistent, incomplete or open to interpretation, AI isn’t going to magically solve those problems.

It may simply reproduce them.

Getting to a useful first draft

One of the things I’ve found interesting is how much time can be saved simply by getting closer to the right result from the beginning.

Think about a typical marketing request.

Someone needs to create a presentation, an email campaign or a series of social media assets.

Traditionally, they’ll find a template, refer to the guidelines, gather the correct assets and start putting something together.

With AI, much of that initial production can happen considerably faster.

But speed alone isn’t particularly useful if someone then needs to spend hours correcting the output.

The real opportunity is getting AI to produce something that’s already working within the brand’s established rules.

That doesn’t mean the first result will always be perfect.

And it certainly doesn’t mean that designers are no longer required.

It simply means we can spend less time correcting basic inconsistencies and more time improving the actual work.

But who approves what AI produces?

This is another area I’ve been exploring.

If AI makes it easier for people across an organisation to create branded content, it also becomes easier to produce considerably more of it.

Which raises an important question.

Who’s checking everything?

Brand governance has traditionally relied on approval processes, design reviews and people knowing when to involve the right teams.

Those responsibilities don’t suddenly disappear because AI is involved.

If anything, they become more important.

That’s why I’ve been working on an Approval Hub as part of the wider system.

The idea is to give teams somewhere to submit their work for review.

An administrator or designated brand owner can approve or reject submissions, creating a central record of what’s been accepted.

Approved materials can then become reusable references for future work.

It’s a relatively simple principle, but an important one.

AI can help create the content. People should still be responsible for deciding whether it’s right.

The importance of a proper handover

Another thing I’ve been thinking about is what happens after a brand system has been delivered.

Over the years, I’ve worked on quite a few design systems.

And one of the biggest challenges isn’t necessarily creating them.

It’s making sure people continue using them correctly.

Documentation becomes outdated. New team members arrive. Different departments develop their own ways of working.

Eventually, the original system can become fragmented.

AI-readable guidelines introduce an interesting opportunity here.

Rather than handing over a collection of documents and hoping everyone follows them, organisations could maintain a central set of instructions that their AI tools reference.

When something changes, those instructions can be updated.

The intention is that everyone continues working from the same foundation.

Of course, that requires proper version control, governance and ownership.

But those are challenges worth solving.

I don’t think this replaces designers

Whenever AI and design are mentioned in the same conversation, there’s inevitably a discussion about whether one is going to replace the other.

Personally, I think that’s missing the more interesting point.

AI is becoming another way of producing digital experiences and creative work.

That means the systems supporting those experiences need to evolve.

Designers still need to make decisions about usability, accessibility, visual communication, brand expression and whether something actually works for the people using it.

AI doesn’t remove those responsibilities.

What it can do is change how we apply the decisions we’ve already made.

And that’s where I think this becomes particularly interesting.

Brand guidelines need to become something we can use

I don’t think traditional brand guidelines are going to disappear.

There’s still considerable value in documenting the thinking behind a brand, explaining its personality and showing examples of how it should be expressed.

But I do think their role needs to evolve.

We’re moving towards a world where designers, marketers, developers and increasingly AI tools all need to work from the same brand foundations.

A PDF sitting somewhere on a shared drive probably isn’t going to be enough.

The more interesting question is whether we can create systems that make it easier for everyone, including AI, to produce consistent, high-quality work.

And perhaps that’s a useful way to test whether a brand system is doing its job.

If someone needs to create a presentation by Thursday, can they produce something that’s genuinely on-brand without having to become an expert in the guidelines first?

If the answer is yes, then we’re probably heading in the right direction.

Quick check

Project video