Tutorials, tool reviews, and half-formed thoughts on making AI actually useful on the job — mostly from the Azure side of the fence.
Six months in. The boring parts got quieter, the interesting parts got louder, and a few things broke in ways I didn't expect.
A working setup with auth, rate limits, and the gotchas nobody tells you about.
RAG over a messy SharePoint got complicated fast. A short log of what went sideways.
I tried roughly twenty. These are the three still open in a tab on my work laptop.
A small defense of keeping the thinking part. Also a small complaint about demos.
A handful of templates that survived contact with real incidents.
I have fourteen versions of the same prompt. Here's how I started deduping them.
It's almost always "problem-first, AI-sometimes." A short case for boring mental models.