I Asked AI to Design My Brand. This Time, We Built Something Together.
DesignA11y had become a website about a practice I had already moved beyond.
The site had been untouched for a long time. The visual identity felt tired. Much of the resource library belonged to an earlier version of the design landscape. Meanwhile, I was publishing articles on LinkedIn that never made it onto my own website.
My thinking was moving. My website wasn’t.
Inside the refresh
Letting go of outdated content
Bringing LinkedIn articles home
Building the identity through critique
Using the AI A11y Toolkit throughout
Discovering the sensory field
The meaning behind the design language
Rethinking the resource library
Same creative director, different team
What changed my mind about AI
In July, I wrote “I Asked AI To Design My Brand Identity.” The results were underwhelming. AI could explain a design concept much more convincingly than it could make one.
This time, the experience was different.
Working with Codex, I revisited the identity, developed a visual language, brought over missing articles, and reconsidered what deserved to stay. Somewhere in that process, a website refresh became a much more interesting experiment in creative collaboration.
And yes, I need to update my earlier assessment. The AI tools I’m using have become considerably more capable design collaborators.
I also became a more demanding collaborator.
First, admitting the site had gone stale
There is a particular discomfort in auditing your own old work.
You remember why you made each decision. You remember when a resource felt useful. You can explain how the page ended up that way.
None of that means it still belongs.
DesignA11y had accumulated pages about individual tools, plugins, color resources, annotation kits, and workflows. Some ideas still mattered. Some recommendations needed replacing. Some pages had simply outlived their usefulness.
The articles were different. They held the thinking, experiments, and perspective I wanted to preserve.
That became an editorial decision: keep the writing, reassess the recommendations, and give future content a clearer home.
We disabled old pages so they remained recoverable. We simplified the resource structure. We questioned leftover membership and login experiences. We stopped treating every existing page as something the redesign needed to accommodate.
A fresh start required deciding what I was willing to let go.
Catching my website up with my writing
I had also created a familiar publishing problem: LinkedIn had become the place where the work lived.
Articles went out through the newsletter. Updating the website became another task to do later.
Later became a backlog.
We used browser automation to help move the articles into Squarespace, bringing over content and cover images and working to retain the original publication dates. I reviewed and published the important pieces.
This was a catch-up workflow rather than a permanent automatic sync. But it removed a significant amount of repetitive work.
It also prompted a bigger question: why was my own website the least current record of my thinking?
The refreshed DesignA11y needed to become a useful destination in its own right.
The identity emerged through critique
The new identity was a true collaboration.
AI proposed the initial open-frame direction: a crisp rectangular outline with a corner extending beyond its boundary. Together, we developed it into a wordmark and a broader design language.
I pushed on the details. All caps. No space between DESIGN and A11Y. A different dark tone. More precise spacing between the Y and the corner. A footer mark that actually matched the header.
Then I pushed on the system.
How did the corner relate to the background? Why did a graphic belong in that section? Were the line weights deliberate? Did an interaction feel integrated, or did it look like someone had found an interesting effect and dropped it onto the page?
There were plenty of rejected directions.
Too chunky. Too literal. Too disconnected. Too much visual interference. Interesting movement attached to the wrong shape.
The useful part was being able to keep responding to something visible.
I could point to what wasn’t working, articulate why, and explore the next version. The concepts and the implementation evolved together.
That felt much closer to a design process than my earlier experiment had.
Accessibility was part of the brief throughout
I repeatedly asked Codex to consult my AI A11y Toolkit when making design and build decisions.
The toolkit provided concrete guidance for contrast, keyboard interaction, focus visibility, motion preferences, semantic structure, and responsive behavior.
Those requirements shaped small decisions with real consequences.
The blog image and title should form one link, rather than create duplicate keyboard stops. Article links need descriptive names. A decorative graphic shouldn’t become screen-reader noise. A pearlescent focus treatment still needs reliable contrast and enough breathing room around the focused element.
Motion needed the same care. Reduced-motion preferences mattered, and the default experience also needed restraint.
Having the toolkit available did not make every first attempt correct. I still had to catch problems, question choices, and ask for revisions. It gave the collaboration a reference point we could return to.
The site still needs ongoing testing, including with assistive technology. Guidance is part of the process; it doesn’t certify the result.
The most interesting discovery wasn’t in the original brief
I wanted the site to demonstrate that accessible design can be beautiful, layered, joyful, interactive, and elegant.
I kept describing the feeling I wanted as an “ADHD brain massage.”
That was a creative brief about my own sensory preferences, not a claim about what every neurodivergent person needs.
We explored backgrounds, reactive graphics, fragments, and movement. Some versions felt disorienting. Others felt arbitrary. Several looked like effects in search of a reason to exist.
Eventually, I began referencing physical sensory objects: elevated adult fidgets, tactile rings, articulated forms, and the satisfying repetition of small elements.
That was where I defined the sensory-field concept.
I didn’t arrive with a finished idea and ask AI to execute it. The idea emerged through looking, rejecting, describing, and making connections to things outside web design.
We moved toward a luminous background with a layered veil of small fragments. Depth came from the image, variation in scale, and selective blur. The fragments could respond delicately without carrying the entire visual experience.
Performance became part of the design critique too. A laggy interaction wasn’t going to deliver the feeling I wanted.
The result came from my sensory references and creative direction, developed through the alternatives we explored together.
A visual language with a reason to exist
The system gradually found its relationships.
The open frame and detached corner express designing beyond the default: questioning the boundary rather than assuming everyone fits inside it.
The small corner fragments carry that geometry into a looser, more organic arrangement. They connect the precision of the frame to the softness of the surrounding atmosphere.
The seam offers a quieter expression of that dispersal within a section, where a line can soften into individual elements.
The pearlescent backgrounds introduce depth and softness alongside bold typography and precise edges. Light and dark variations let that atmosphere appear across the site without repeating the same composition everywhere.
The sensory field gives that language a more immersive, responsive expression.
Aubergine, blush, copper, and warm light connect the different treatments. The identity can be structured in one place and atmospheric in another while retaining a recognizable relationship.
That relationship mattered more to me than adding another impressive graphic.
The resource library needed a new point of view
When I first built DesignA11y, a collection of Figma plugins could represent a substantial part of an accessibility-minded design workflow.
Plugins haven’t disappeared. But they no longer describe the whole landscape—or the questions I’m exploring.
The refreshed library reaches into AI-assisted workflows, machine-readable annotations, design-to-code context, accessibility guidance for AI tools, design systems, inclusive content creation, representation in generated imagery, and responsible motion.
The selection criteria changed too.
Does this help someone make a better decision? Is the guidance current? What does the tool actually do? What still requires human judgment or testing?
A resource library needs an editorial position. Otherwise, it becomes an inventory that quietly ages.
I want DesignA11y to be a resource for inclusive design as it is practiced now, with room for the tools, skills, and questions that come next.
My role felt surprisingly familiar
As an Executive Creative Director, I recognized how much of my role remained the same.
I was defining the ambition, bringing references, evaluating concepts, questioning whether the elements belonged together, and pushing the work until it felt intentional. I was connecting individual decisions to a larger story.
Those are familiar responsibilities from a much more traditional creative process.
What looked different was my team—and how quickly we could move between an idea, a visual exploration, and a working implementation.
That distinction feels relevant as our practice shifts. Many forms of execution are changing substantially. In this project, my role stayed remarkably recognizable.
I still needed to articulate what mattered, recognize potential, give useful feedback, and know when something wasn’t good enough.
I also had to make that feedback actionable. “This isn’t landing” could start a conversation, but progress came from explaining the relationship that was missing, the feeling I was after, or the detail that broke the composition.
The tools expanded what I could explore and execute. My experience helped me decide what was worth pursuing.
What changed my mind
My earlier experiment exposed a gap between AI’s ability to describe identity design and its ability to do it convincingly.
This project gave me evidence of a different kind of capability.
AI contributed concepts, visual exploration, implementation, and the repetitive work that had kept the site behind. I contributed the purpose, references, critique, sensory perspective, and decisions about what belonged.
Neither contribution was incidental.
The strongest moments happened when an output helped me recognize an idea I hadn’t fully articulated yet—and I could then give it direction.
DesignA11y now has a clearer identity, a more considered resource library, and a home for writing that had been scattered across platforms.
It also has a more useful role: a place to keep exploring how inclusive design and AI can work together, with the process visible.
Including the versions that didn’t work.
