Designing an AI supervisor that clinicians would trust with a session

We designed the review as two layers: skill-level feedback underneath a whole-session evaluation, so a supervisor sees the pattern before the detail. The identity reads as clinical rather than consumer AI, which is what a HIPAA context demands, and the competency view had to stay legible to a director tracking many supervisees at once.

• Earn a clinician’s trust before asking for a recording.
• Show development over time, not a score per session.
• Serve nine therapeutic approaches without a separate interface for each.

An AI that listens to therapy sessions starts from a trust deficit, and the design had to answer that before it sold anything. Two audiences share one screen: the supervisor reviewing and the supervisee being reviewed. The product was early enough that the brand had to hold while the feature set moved underneath it.

One interface carries all nine therapeutic approaches. The competency view reads at a glance rather than needing interpretation, and early access runs on the product and the pitch we built.