Insights

What actually happens on AI workflow builds at small and mid-sized companies, and the principles that came out of them.

2026-10-22

All the tests passed. What else should you check?

A passing result only counts as evidence when you know what it verified. Three moments when I nearly said it was done, and four questions to ask before trusting that everything passed.

2026-10-08

You have all the client’s files. Why is the thing that matters still missing?

What is missing never announces itself. Four moments on one project where I missed a gap, and the four questions I now ask every time.

2026-09-24

How do you tell which work should not be automated?

The tasks that look like perfect candidates are sometimes the ones to cut. Four criteria I actually use to take work off the list.

2026-09-10

Why does the work take just as long, with five AI tools open?

A company paying for five AI tools still spends a week on a single product page. Three real cases explain why.