AI in Clinical Supervision: Practice and Human Connection

In the current climate, concerns around AI and how it will change us and our world are understandably high. What happens to human relationships when AI begins to enter so many facets of our lives and risks becoming a barrier to connection instead of a facilitator? In my own work, I wonder what happens to the importance of human-human interaction when AI is introduced not only into various aspects of clinical work, but into the supervisory process as well. Perhaps, though, the question isn’t the knee-jerk one of whether AI belongs in supervision, but what we ask it to do.

When we think about introducing a new technology, technique, or facet to something as significant as supervision, it’s important to reflect on what we are doing and how it may change things. What are we asking it to do? How will it affect outcomes? What may be lost by changing the way we do things? Just because something is cheaper, faster, or newer doesn’t make it better. Integrating AI into supervision therefore requires us to consider what it can do in and for supervision.

New clinicians often require practice, on-demand support, and a way to develop their skills without risking client welfare. Some of those needs are things AI can assist with, but not replace a person. I think that distinction is crucial. AI can simulate a clinical situation, offer considerations for decisions, and provide general feedback. It can’t take responsibility for clinical decisions or replicate human connection. Asking it to do so would be like trying to replicate equine therapy with a motorcycle. It can help move a person from A to B, but it’s not doing much for coregulation.

This distinction helps us begin to look at what supervision is, and not just what it accomplishes.

Supervision is so much more than quality assurance, teaching interventions, reviewing documentation, and skill acquisition. It certainly includes all those elements, but good supervision moves far beyond them. A good supervisor recognizes the importance of human connection alongside the important content of training. How can a clinician respond to an angry client? How can one work through their own feelings of inadequacy or ineffectiveness? What does it feel like to stay in discomfort that is productive? How can one differentiate between clean and dirty pain? When is uncertainty helpful? These are not simple situations with quick answers. They are the kinds of questions that require human connection and lived experience to engage.

A good teacher absolutely helps their student grasp the course content. But they don’t stop there. The teachers who stay with us and continue to guide us long after a class has finished are the ones who modeled and lived out their teaching. They recognized the formational significance of their role. As clinicians, we know all too well that change is not merely a matter of information. If it were, TED Talks and self-help books could replace therapy. Change is far more complicated, human, and nuanced, and the same is true for our supervisees.

So where can AI support supervision?

Like a calculator, word processor, or EHR, AI is a tool that can do certain tasks efficiently. AI can be really helpful between supervision sessions when a supervisee wants to work on a new intervention, practice with a new population, or review their clinical judgment. Specifically, I believe AI can help with Practice, Reflection, and Feedback.

In practice, supervisees can rehearse difficult sessions and clinical situations repeatedly without waiting for them to arise in clinical practice. AI can also help clinicians examine a case from multiple angles, different theoretical approaches, and various cultural perspectives. What am I missing? What assumptions am I making? How else might the client experience this intervention? It can be used to facilitate growth in their capacity for cognitive complexity.

Finally, AI can be really helpful in providing immediate, low-stakes feedback on issues like documentation, treatment planning, and simulated clinical interactions. AI can make practice more available, not supervision less relational.

And here is the key: it is so important that supervisors recognize what AI can do well so that they don’t try to force it into roles for which it is not well suited. Don’t outsource the supervisory relationship.

AI can handle very particular tasks well. It can’t replace you. Wikipedia did not make textbooks or teachers obsolete.

To that end, as we take cautious steps forward into new technologies and approaches, it’s important to remember a few things.

AI should never:

Replace a human supervisor
Make final clinical decisions
Substitute for relational attunement
Become an escape from the discomfort of bringing something difficult to an actual supervisor

Instead, AI should be used in service of creating better material for human supervision.

The goal isn’t to make supervision less human. It’s to wisely use tools like AI to create more opportunities for clinicians to practice, reflect, fail, and try again allowing that the precious time they spend with a supervisor to go deeper.

Systems like PraxPlay aren’t trying to make AI your supervisor; it's trying to give you a better place to practice before you meet with one.