How I work with AI

From the initial idea to review

I enjoy creating and developing concepts. I define what I want to communicate, who it is for and why.
The ideas and the creative and editorial direction come from me. AI helps me develop them, explore possibilities and automate tasks.
I do not settle for its first suggestions. I discard, rephrase and refine them until I recognise my voice in the result.
I keep improving the Atelier’s design and user experience through what I learn and feedback from the people who visit.

I prepare and refine my instructions with help from AI, through metaprompting.
I look for reliable sources and cross-check data and references. I distinguish data from my interpretations and identify synthetic material.
I turn to other AI tools to compare responses and identify possible errors. I decide what needs further review.
I review inclusive language, readability, image alternative text and ease of use.

I decide what to share and take editorial responsibility for what I publish.
I explore which tasks to delegate to a machine and when to stop, review or decide.
I care about how a result may affect other people and which decisions need human review.
Before introducing AI into a professional project, I consider it a priority to agree on the permitted data and tools and who will review the results. This includes how access, retention and deletion of information will be managed.

Behind the scenes

Connections can change and lead me back to review an earlier stage.

I work with ChatGPT Work and Claude Code on my project files. APIs and MCP let me connect external applications.

With each trial, I check what works, what errors appear and what needs adjusting before I continue.