Design in AI-First Marketing

The AI Journey at Forge

The use of AI in the design workflow at Forge has been a journey, starting from a place of extreme reservation, then developing into a workflow we can confidently apply in the areas we deemed most impactful.

Isn’t AI design just theft and slop?

When I was first approached about incorporating AI into the design workflow at Forge, I was extremely sceptical. Up until that point, all the AI “art” I had seen could comfortably be labelled as theft and slop.

Yet, with AI spreading like wildfire, it quite quickly became apparent we would have to confront its implication on design. This went hand in hand with its expansion into the other parts of the business, as well as our increasing familiarity with some of our clients' own complex systems, helping to highlight the potential.

How can AI make accurate and consistent designs?

That’s where I started learning about Retrieval-Augmented Generation (RAG), a method when the LLM pulls information from specific internal documents to have better control and reliability of what the model produces; I began to understand how it could be useful to us.

One of the first use cases was for a client we had already produced a significant body of social content for, where we would spend time each week updating copy and layouts for consistent templates. The model was able to take these designs and reproduce them effectively, meaning we could free up more time to take future carousels in new directions.

This also goes hand in hand with the ubiquity of AI across design tools, like figma make, and the speed of improvements in Claude design, which meant when used under strict rules could match our pre-existing designs.

Now we have the ability to take the copious designs we’ve worked on across our clientele and develop methods of automating repetitive work, while still maintaining complete creative control and ensuring all our work has initially been designed by hand, in-house.

It’s been very important to us at Forge to not just keep up with the demands of the current AI ecosystem, but to do so with as much integrity as the space allows. Which means strictly using our own work and expertise to fuel our own AI knowledge base, and ultimately produce work of both quality and speed that we can be proud of.

What has it changed for our team and clients?

One of the biggest benefits the knowledge base has created is the massive improvement in velocity and capacity across the company. Team members outside of the design team can now quickly and effortlessly have their work designed, allowing them to see a closer resemblance to the final asset much sooner.

This still culminates with an ultimate review and edit by the design team to ensure the branding and rules are being adhered to, but has relieved some of the menial work, and allowed the team to spend more time focusing on the creation of new designs and assets.

By implementing this workflow throughout our client base, we’ve been able to cut back on the usual hours we spend making reliable assets like social content, and take on significantly bigger tasks like rebrands and website redesigns to create additional value for our clients. It also means design can keep pace with our GTM & strategy work, where launches and campaigns need new assets quickly.

Doesn’t AI-generated design all look the same?

Ultimately, the impact of velocity and capacity is a double-edged sword, as it has also meant that many non-designers have the ability to quickly generate new designs they previously wouldn’t have been able to. This is something we often see outside of the company, where assets are readily being pumped out that are more often than not of questionable quality and impossible to differentiate from any other AI-generated work.

That’s why it’s a foundational belief of ours that for AI to be used effectively, it has to be built off of our pre-existing work, grounded in solid design principles, and must always go through human review by the team.

If you’d like to see how this could work for your brand, book a call.