AI as an assistant, not replacement
Let me say the most important thing first: AI is not a therapist, and using it as one is dangerous. Real therapeutic work requires a licensed human being who can hold complexity, notice what you cannot, and stay with you over time. Nothing in this post replaces that.
What I want to share is something different — a workflow that took a static resource my therapist gave me and turned it into an active week of practice, while keeping every clinical judgment in human hands.
The Situation
My therapist gave me a self-paced chapter on a topic we’re working on together. It was well-designed: a chapter to read, a set of frameworks, and a fill-in worksheet. Most clients, myself included, read this kind of material once and shelve it. The resource is solid. The friction is in turning a chapter into daily practice between sessions.
This is where AI earns its place — not as a clinician, but as a structuring tool.
The Three-Step Workflow
1. Restructure the resource into a multi-day format.
I uploaded the chapter to an AI assistant and asked it to build a six-day companion guide drawn entirely from the resource’s own language. The instruction was specific: pull from this text only, do not invent new frameworks, and surface the items from the original most relevant to my work. The output was a workbook in the same voice as the original — same concepts, same framing, broken into daily practice.
The key move is constraint. AI was not asked to be creative. It was asked to reorganize a human-authored resource into a usable format.
2. Do the work each day, by hand.
The daily practice itself was unassisted. Read the day’s section. Sit with it. Write the responses. No AI in the room for this part — the work has to be yours or it doesn’t count. The thinking is the practice. It cannot be skipped.
3. Use AI as a reflection partner on what you wrote.
After completing each day’s entries, I shared what I had written with the assistant and asked for reflection — patterns it noticed, places my writing contradicted itself, language worth holding onto. Three rules governed this step: no diagnostic language, no rewriting my words into something I didn’t say, and surface rather than solve. Anything that needed solving was set aside for my next session.
Why This Matters for Leaders
The same pattern applies to any human-authored resource your people aren’t actually using — training materials, frameworks, coaching workbooks. The bottleneck is rarely the quality of the resource. It’s the friction between a one-time read and daily practice.
AI is exceptionally good at reducing that friction when given three things: a strong source document, a tight constraint on what it’s allowed to do, and a human owning every judgment that matters.
The Failure Mode Worth Naming
The danger is not that AI is too weak to help. It’s that AI is too willing to do whatever you ask. Ask it to be a therapist, and it will perform one. The discipline is in the prompt: define narrowly what AI may do, and refuse anything outside that scope — even when it offers.
Human connection is not a limitation to work around. It is the part of the system that gives the rest of it meaning. AI is a tool that makes the human work go further. Used the other way around, it becomes a substitute for the work — and the work is the point.
What human-authored resource — a book, a framework, a coaching tool — have you read once and shelved, and what would it look like to actually build a practice around it?