Picking a Brand Look as a Non-Designer: AI Made Prototypes Cheap, So Taste Became the Hard Part

I am not a designer, but I spent this week iterating on a visual system for AgentCrew Academy with Fable, GPT-image-2, and GPT-5.6-sol. Here are the four rejected directions and the final spec applied to the website, slides, and documents.

I’m not a designer, but I spent a good chunk of this week picking a visual system for AgentCrew Academy.

I started by having Fable draft a proposal, and the first version already caught my eye: white, ink, vermilion, black. Quiet, but decisive. This is a screenshot of the original Fable artifact, still openable, not something reconstructed after the fact.

A Fable artifact screenshot showing the first visual proposal, a white-ink-vermilion-black palette paired with the tagline "quiet, but decisive," applied to an AgentCrew Academy website prototype

One version wasn’t enough, so I asked it to go bolder. That produced four rejected directions, A through D: gallery white, warm-paper editorial, ink-black cover, navy professional. Each one pulled from a different reference. NENDO’s typography summer school. The warm-paper editorial feel of &PREMIUM and Kinfolk. The ink-black covers of BRUTUS and SUPPOSE. The navy professionalism of Takram and SmartHR.

Four rejected visual directions, A through D, shown side by side: gallery white, warm-paper editorial, ink-black cover, navy professional

Later I tried a completely different approach, skipping Fable this time and asking GPT-image-2 directly for 5-10 style variations to choose from. Then I fed my pick back to GPT-5.6-sol and asked it to mock up a YouTube thumbnail. I also gave it my current YouTube channel homepage and had it build an HTML prototype of what the redesigned channel would look like. Once I settled on one, it built me a thumbnail template so I could swap in new content and titles going forward. I haven’t actually switched the channel art yet. I want to run an A/B test for a while first.

Working with GPT this time around felt just as good, so I think either path works. In the AI era, producing a prototype barely costs any time anymore. The hard part has shifted back to your own taste and judgment.

What I finally settled on wasn’t a moodboard. It was one visual language rendered into three real outputs: the website, the workshop slides, and the A4 documents are all HTML-rendered, so changing the style still shares the same set of design tokens.

Screenshots of the same visual language applied side by side to the website, an online workshop slide deck, and an A4 document for an HR enterprise-adoption case study

Along the way I also tested an amber variant, comparing it across the website, a YouTube EP.32 thumbnail, and the HR case-study document.

An amber-toned website variant, a YouTube EP.32 thumbnail, and an HR adoption document shown side by side to compare the same visual language applied differently

I’m planning to package this whole “picking brand visuals as a non-designer” process into a skill and release it later.