How I Use AI
Short version: my work is human-first. I use AI as an assistant, not a replacement. Here is what that means in practice.
The tools
Mostly Gemini and Claude for words, and Canva AI (and sometimes others) for images. That list changes, so I'll update this page when it does.
Where this applies
Everything I publish: the articles on this site, my weekly Focus. #deliver newsletter, and the podcast.
What I use them for
- Brainstorming. Angles, titles, structure, the question I hadn't asked yet.
- Drafting and editing. It depends on the piece. Some are mine start to finish. Some start as my notes or ideas that AI helps shape. Some start as an AI draft that I then rewrite heavily, in my voice, with my examples.
- Transcribing and organizing. Turning audio into text and formatting my rough notes.
- Images. I often use Canva AI, and sometimes other tools, to create the background image for a new article.
- Research and summaries. Reading, summarizing and pointing me toward sources I then look at myself.
- This website. Claude helps me write and maintain the site's code and tooling.
- The podcast. Implementing Scrum: Unscripted is human created, with AI-generated voices, mostly from NotebookLM (now Gemini Notebook). They are not clones of real people. It is an ongoing experiment, and it says so on its page.
What stays human
- The experience. Stories, client examples and lessons come from my own work as a Scrum trainer and practitioner. I don't ask AI to invent them, and I don't publish invented ones.
- No made-up people or quotes. No AI personas, no fabricated testimonials, no quotes attributed to people who didn't say them.
- No voice cloning. I don't clone anyone's voice. The podcast voices are generated, not copied from a person.
- No AI-made images passed off as real. I don't present AI-generated images as photographs of real people or events.
- The decision to publish. My name is on this site and I approve what goes on it.
Labels
If a piece is mostly AI-written with little change from me, I'll say so in a note at the top.
Private and client information
When the material is confidential or about a client, I use private or local AI models. When I use a public tool, I turn off training on my data and limit what the tool can access.
How I check the work
I review what I publish, but I'll be honest: I don't always read every line of every piece the same way. AI can get facts wrong, and so can I. If you spot something wrong, I want to know.
Found a mistake?
Please tell me which page and what's wrong. I'll fix it and, where it matters, note the correction on the page.