I am most interested in AI when it stops being a demo and removes a piece of work I actually have to do.
That has led me to build small internal tools rather than wait for a perfect software product: a scouting assistant that makes thousands of technology providers searchable in natural language, a bot inside Notion to replace manual filtering, and a lightweight time tracker for workshops.
Start with the annoying task
My rule is simple: I do not start from “where can we use AI?” I start from something repetitive, slow or difficult to search.
Then I ask whether AI, automation or a small no-code tool can make that task meaningfully easier. Sometimes the answer is yes. Sometimes a spreadsheet is still better.
Small tools compound
Internal tools rarely look impressive from the outside. That is fine. A useful tool can save minutes every time a process repeats, make knowledge easier to retrieve or remove a dependency on one person remembering how something works.
The scouting system is the clearest example: instead of manually navigating a large technology landscape, the goal was to make the underlying knowledge easier to query against a specific business need.
Build before buying — sometimes
I am not arguing that every company should build its own software. Most should not. But modern AI and no-code tools have changed the threshold for experimentation. You can now test an internal workflow before turning it into an IT project.
That is the part I like: getting close enough to the problem to build a rough solution, see whether people actually use it, and only then decide what deserves to become more serious.