Stop looking for a use case — automate something useless first

"Okay, but what should I actually automate?"

Stop looking for a use case. Automate something useless first.

I asked Claude Code to build me a Spotify playlist of every song that was #1 on my birthday — for every year I've been alive, US and German charts. I didn't need it. That was the point. It was the perfect little test of what these tools can actually do, without anything real on the line.

Here's how it went, step by step:

1. I said what I wanted. "The #1 song on each chart, on my birthday, for every year since 1990." (the result that needs to be clear)
2. It did the research and pulled 35 years of chart history, matched a single date to weekly charts, including the years the #1 changed mid-week. It handed me a clean year-by-year table to check. (the data that needs to be clean)
3. It built the playlist in Spotify directly. Connected the tool, created the playlist. No copy-pasting 36 songs by hand. (the action, done for you)
4. It hit errors — and worked around them out loud. Spotify's tool had a song limit per playlist, and one old German track refused to cooperate. Instead of failing quietly, it split the work and told me exactly which song to add by hand. The only thing I did manually was the last step: merging the 5 playlists into one. (that's why we keep the human in the loop for now ;) )

Total effort on my end: a few sentences and a couple of decisions.

That's the whole lesson. You don't need a "use case" to get a feeling for automation. You need a small, slightly pointless task — and suddenly you understand what the tool is good at, how the connectors work, and where it breaks.

A prompt idea for your next office party, if you want to try it yourself:

Create a Spotify playlist for our office Christmas party. Mix classic feel-good Christmas songs with upbeat crowd-pleasers from the last few decades — nothing too slow. Keep it fun and singalong-friendly, around 40 songs.