Automation isn't just a choice where you snap your fingers and it's done. Automation takes time and effort and adds technical debt and support burden; sometimes these are trivial amounts, sometimes they aren't.

I use a service that sends me a bunch of fairly trivial Google Analytics data. It costs $10/month for the equivalent of fifty lines of Python.

I know I could spend maybe two hours writing a replacement for it and set it on an EC2 host and probably never worry about it again, but those two hours are still non-trivial amounts of time: two hours of my time pays for more than three years of that $10/month. Plus, the API may be brittle, there may be edge cases I haven't thought about, and I don't want to add yet another thing to monitor and worry about.

Sometimes, yeah, you can automate things and roll your own dependencies -- and sometimes the calculus works out and it's a no-brainer. But I've found that homebrewing a $XX/mo dependency is rarely the best use of my engineering time.