I Was My First Client.

I have been self-employed since 1999. Photography, design, art. Working like this means doing a great many different things in a single day: writing a quote, sorting pictures, changing a website, an email to an association, an invoice, a press release, an application for an exhibition.

Each of these things had its own program. And each program wanted me to explain from scratch who I am.

It started small

In 2025 I began working with Claude, an AI made by Anthropic. At first for individual tasks: a tool that sets PDFs in my design. One that prepares emails as drafts but never sends them itself. One that builds my website.

The problem quickly turned out to be the same as before: each tool only knew what I had just told it. So I started writing things down. How I speak. Which colours I use. Which words I do not like. What each client has already decided.

Bit by bit, these notes became the most important part of the whole system: a shared memory.

Helpers became a studio

In summer 2026 I put the whole thing in order. There are now rooms: for clients, for design, for the website, for texts, for documents, for research, for rules. Every client has an office with the same compartments. Every task has a fixed responsibility, and I have given these responsibilities names and faces. The copy editor reads every text before it goes out. The fact checker checks figures against the source. The juror looks at design without knowing how it came about.

That sounds playful. But it helps me see who is working right now, and on what.

And then I took things away again

This is the part that is usually missing from stories about AI. In July and August I had my own system reviewed. The result was clear: many rules stood in several places at once, and some duties existed only on paper instead of being enforced. The guideline since then: build less, check more, slim down, secure.

I built a workshop for my post, ran it for a while, and abolished it again. I prefer to sort my emails myself. That was not a defeat. It was the answer to the question I ask of every tool: does it take work off my hands, or does it create new work?

What went wrong

Quite a few things went wrong, too. Once a test sent a real access email to a client before her access was ready. Since then there have been three independent safeguards so that it never happens again. Once a form was quietly broken while every indicator glowed green. It was found by a routine that opens the page the way a visitor would.

Each of these mistakes became a rule. And the rules became an attitude: the machine may prepare a great deal. The decisions are mine.

Why I am passing it on now

At some point I noticed that the businesses I work for have the same problem as I do. Only bigger. More people, more places where knowledge sits. More gaps between them.

That is why Kenala grew out of my studio. What is the shared memory for me is called the DNA of a business there. The website is the first area that grows out of it. Social media, printed matter, and newsletters grow from the same centre.

My own website was the first to be looked after this way: one sentence, one preview, one yes.

This has not turned me into a programmer. I am still someone who designs. I just also design the things that make design hold up in everyday life.