Three Takeaways on AI in Therapy From My CAP Talk

I recently spoke to the College of Alberta Psychologists about using AI in therapy. The talk ran over an hour, so here's the short version: the three takeaways I think matter most. Watch the video recap on YouTube.

1. When your client brings AI into the room

A client tells you they've been uploading your sessions to ChatGPT and talking them over with a chatbot. It feels weird, maybe even like a violation. That's a fair reaction.

Get curious before you get defensive. Find out what they're getting out of it, and think of it more as an interactive journal than a second therapist. Then help them see what they're giving up: those chats don't carry the confidentiality of the therapy room, and those platforms don't protect PHI.

2. Vet the organization when you can't vet the tool

Most therapists aren't that tech savvy. If you're choosing an AI scribe and can't evaluate the technology, evaluate the company. Where does my data live? Will you train on my clients' data? Will you sign a written BAA? If you can't get straight answers, there's your answer.

No matter how cool the tool looks, if something goes wrong, it's your license on the line.

3. What we haven't figured out

Should we use these tools at all, given how they were trained and what they cost? Is opting out even a real choice, when our clients and organizations are adopting them anyway? At minimum we need to know enough to talk about them competently.

And if we get this right, what could practice look like? Tools that let us be present instead of documenting for billing (more in What Comes After the Note). Supervision based on what happens in the room, not what we remember a month later. Client tools built around their lives instead of a photocopied handout.

Tools aren't inherently safe or unsafe. How we use them, and how we evaluate them, can be. If this sparks something for you, I'd like to hear it.