I have worked with Wendy for a long time. For years, she ran an online retail business, and I helped with the website, ecommerce, marketing and all of the usual things that go along with running a business online.
Recently, Wendy decided to take the business in a very different direction. The new business is much more personal. It is built around her art, her jewelry, the stories of women, and some fairly deep ideas about change, identity and where women find themselves at different stages of their lives.
That presented me with an interesting problem.
I could build the website. That part wasn’t particularly difficult. What I wasn’t sure I could do was design a website that actually felt like Wendy.
The Part of Web Design I Have Always Hated
For me, one of the hardest parts of website design has never been WordPress, CSS or programming. It has been trying to get inside somebody else’s head.
A client might tell you they want something elegant, artistic, feminine, modern or warm. Those words can mean completely different things to different people.
You build something, show it to the client and wait.
Sometimes they love it. Sometimes they hate it. More often, they say something like, “I don’t know what it is, but it doesn’t feel right.”
That feedback can be frustrating for a designer, but the client isn’t wrong. They are reacting to something they can feel even if they can’t necessarily describe it in technical design language.
With Wendy, I wanted to see if AI could help solve that problem differently.
What If the AI Could Learn Wendy First?
Instead of simply asking AI to “design an artistic website for a jewelry business,” I started by giving the AI context about Wendy.
I already had an advantage because I have known her and worked with her for years. I understand quite a bit about how she approaches her business, what matters to her and what she is trying to accomplish with this new direction.
We also got extremely lucky.
Wendy has written extensively over the years. She has blog posts where she talks about her work, her customers, creativity, women, personal change and many of the ideas behind this new business.
That material was incredibly valuable because it wasn’t marketing copy written by somebody else. It was Wendy explaining what Wendy thinks.
So I started giving the AI that context.
Then I gave it additional skills related to website design, user experience, visual direction and the technical requirements of actually building the site.
At some point I realized that I wasn’t really asking AI to design a website anymore.
I was building a kind of virtual Wendy website designer.
The Virtual Wendy
The idea is actually pretty simple.
The AI has one set of information that teaches it how to design a good website. It has another set of information that explains the business, its goals and what the website needs to accomplish.
Then it has something much more unusual.
It has context about Wendy.
Her writing. Her preferences. Her philosophy. The things she likes. The things she doesn’t like. The kind of emotional response she wants the site to create.
The goal isn’t to pretend that the AI is Wendy. It is to give the AI enough context that when it has to make a design decision, it can ask something close to:
What would Wendy probably want here?
That turned out to be much more interesting than I expected.
Then We Let It Design
We used the system to create several initial design concepts.
This wasn’t a case of generating an AI image and calling it a website. These were actual website concepts that could become the foundation of the production site.
I sent them to Wendy expecting the normal design process to begin.
Her initial response was:
“OMG!!!!! Thank you so much. I love this direction.”
That was a pretty good start.
The site is still in production, so I am not going to pretend this is a finished case study with a neat before-and-after result. There is still plenty of work to do.
What happened next, though, is actually the part I find most interesting.
Wendy Started Training Virtual Wendy
Once Wendy had time to really look at the concepts, she started giving me more detailed feedback.
She liked one concept much more than the others. She liked its quiet feeling and the almost monochromatic approach. She had opinions about the font. She wanted certain colours available in the palette, but she didn’t want those colours overpowering the design.
She liked burnt sienna. She liked plum, ochre and gray-green. She wanted teal available too, but used carefully.
That feedback didn’t just become a list of website changes for me to make.
It became new context for the AI.
In practical terms, the real Wendy was teaching the virtual Wendy.
The next time the AI works on the website, it doesn’t simply know that Wendy likes a particular design. It knows more about why she likes it.
That distinction matters.
The system gradually becomes better at making decisions that fit her taste without me manually redesigning every section each time she gives feedback.
That’s the Learning Loop
This is the part of AI that I think many businesses are still missing.
The interesting opportunity isn’t simply asking ChatGPT a question and getting an answer.
It is creating a system where AI has the right skills, the right business context and a way to learn from human feedback.
In this case, the loop looks something like this:
Learn about Wendy → create a design → get Wendy’s feedback → update what the AI knows → create a better design → repeat.
No, we aren’t retraining an AI model from scratch every time Wendy sends me an email.
We are improving the context, rules and knowledge the AI uses to make its next decision.
From the client’s perspective, though, the effect is very similar. The website designer gets to know her better every time she gives feedback.
This Is Bigger Than Website Design
The more I work with AI this way, the less interested I am in the question, “What can ChatGPT do?”
I think the more useful question for a business is:
What would become possible if AI understood how your business actually works and how you make decisions?
In Wendy’s case, we created something that helps make website design decisions in a way that reflects her taste and philosophy.
But the same idea can apply almost anywhere.
You could build an AI assistant that understands how your company qualifies leads. You could create one that knows how you review marketing campaigns. You could create one that understands how an estimator approaches a quotation, how an operations manager reviews an exception, or how a business owner evaluates an opportunity.
The value isn’t just giving AI a task.
The value is giving AI skills, context and a feedback loop.
That is when it starts becoming something more useful than a chatbot.
And We Are Still Learning
I don’t know yet exactly where this particular experiment will lead.
The website isn’t finished, and I fully expect Wendy to give me more feedback before it is.
The difference is that I’m actually looking forward to that feedback.
Every comment gives us more information about what she likes, what she dislikes and how she thinks about the business. Instead of every revision simply creating more design work, every revision makes the system better.
That might be the most interesting thing I have learned from the project so far.
I didn’t just use AI to help build Wendy a website.
I used AI to help build a website designer that is gradually learning how Wendy would build it herself.
If you’d like me to build a virtual “you” to build your website or perform any other task you’d rather AI would do for you, please reach out.
