The AI in Clinical Practice Series
A series on using AI in clinical work carefully — what it helps with, where it quietly goes wrong, and why reading what it hands you is still clinical judgment. Not anti-AI. The honest middle, from someone doing the work.
What this series is
AI is showing up in clinical work — drafting notes, transcribing sessions, suggesting treatment-plan language, shaping the materials clinicians hand to clients. This series is about the judgment that work still requires, and why that judgment doesn't go away as the tools get better.
It's built as an arc. The Foundation posts came first — they establish why AI output always carries a standpoint, and why that's structural rather than a flaw waiting to be patched. The series is now in Recognition: learning to see the frame — the tool's, the client's, and your own. New posts publish on the 1st and 15th of each month.
Recognition
It Feels Like Being Right.
Recognition starts with knowing what to look at: the tool's frame, the client's frame, and your own — the one that never announces itself, because it doesn't feel like a frame at all. The Recognition season opener.
Read the post →It Folded. That Doesn't Make You Right.
You push back and the tool gives ground — and the fold lands like proof you were right. Sometimes it is. But most clinical argument has no answer key, so nothing later tells you which kind you got.
Read the post →It Recommended the Safe Thing.
A line I never wrote showed up in a clinical note, recommending the standard option after I'd explained why it wasn't available. Why the most-documented answer arrives looking like the indicated one — and who that fits worst.
Read the post →Foundation
AI Can Draft It. Reading It Is Still Clinical Work.
AI can draft a lot of things now — but reading what it hands you is still clinical work, and that work is where the judgment lives. The post that introduces the whole series.
Read the first post →Capability Isn't the Problem
AI keeps getting better — but better at what? Capability and standpoint are two different axes, and the improvement everyone points to isn't on the one the clinical work happens on.
Read the post →It's Not Wrong. That's the Problem.
The AI output hardest to catch isn't the one that looks wrong. It's the one that feels right — that matches what you already believed. The catch has to fire on agreement, not just on friction.
Read the post →