Slack Agent Side Quest

I solo built an AI Slack agent for my team this year and didn’t write a single line of code. Before AI that sentence would’ve been an oxymoron. I don’t code. I literally cannot code. A year ago, if I wanted internal tooling like this, I would’ve had to convince a leadership team to allow me to pull an engineer off a higher-priority roadmap item to build a side quest internal tool for a PM. In the before times that would have been a non-starter, and so the idea would’ve stayed on the shelf.

This one didn’t stay on the shelf. It reads our inbound channels, dedupes tickets against our issue tracker, and files new ones with full context and a proposed priority. It classifies reactions on messages and acts on them. About thirty people use it now, including other product managers who run it against their own backlogs. The feature list kept growing because every week I’d notice another seam where the manual version was eating our time, and I could just go build the fix.

The one part I’m proud of technically is the migration. The first version ran on the Claude Agent SDK locally on my laptop, which meant it broke whenever my laptop slept, whenever my Wi-Fi blinked, whenever I closed the terminal by accident. I moved it to Claude’s managed agents API and stood up the Slack webhook gateway on Google Cloud Run. It has been running without me touching it since. None of that work would’ve been possible without being able to ask Claude questions at every step, and trust me, I had a lot of questions.

That’s the thing that’s cool to me. Not the features. The fact that a year ago this was an idea I would’ve filed away because I couldn’t build it and couldn’t ask for it, and now it’s live in my team’s Slack helping us move faster. I know I didn’t build anything too crazy, but being able to build entirely on my own is a new found experience that I’m excited to explore further.