Simulation & VR
Training mental health crisis intervention with AI: from suicide risk to de-escalation
The stakes
The conversation no one wants to handle unprepared
A professional who has never practised asking about suicidal ideation is an institutional risk, not just an individual one. These are the exchanges where the right phrasing, the right pause, the willingness to ask the direct question all decide the outcome — and they are precisely the ones traditional training leaves to chance, because you cannot rehearse them on a real patient.
Why AI
Why role-play with actors does not scale here
Standardized patients are excellent and scarce. A trained actor can run a handful of encounters a day; a service that needs every nurse, resident and psychologist to rehearse a suicide-risk assessment several times cannot wait in that queue. Conversational AI removes the bottleneck: the patient is available at any hour, stays in character, and the conversation can be repeated until the approach is sound — at no risk and no actor cost.
You want the first time a professional asks "are you thinking of ending your life?" to be a rehearsal, not a real patient.
In the field
How a public health service trains this today
This is not hypothetical. The Servicio Andaluz de Salud runs this exact training across two Andalusian hospitals: resident doctors in gynaecology and paediatrics, nurses and psychiatry staff rehearse the clinical interview for suicide risk, acute psychosis and violent behaviour with conversational AI patients, in a hybrid format alongside their clinical work. The full deployment is documented in the case study.
Two details of that programme are worth copying. First, the scenarios are built around the scales professionals already use — SAD PERSONS, Columbia — so the rehearsal reinforces the protocol instead of competing with it. Second, the training targets mixed profiles at once: the point of a crisis interview is that whoever is in the room can hold it, not only the psychiatrist on call.
The session
What one rehearsal actually looks like
A session is short and specific. The professional opens the case, reads the context — a patient admitted after an overdose, a family member escalating in a corridor — and starts talking. The AI patient answers in character: evasive, irritable, silent when the question is clumsy. There is no script to follow and no observer in the room.
The value arrives afterwards. The conversation is recorded and transcribed, and the professional sees where they hesitated, where they softened the direct question into uselessness, where they missed the opening the patient offered. Then they run it again. Confidence in these conversations is not built by knowing the theory — it is built by having already survived ten versions of the moment.
In practice
Three protocols you can train
- Suicide-risk detection: asking directly, assessing from the conversation, building a safety plan.
- Verbal de-escalation: reading the escalation and lowering tension before a situation becomes unsafe.
- Communication with families: grief, breaking difficult news, and conversations under high emotion.
For hospitals · Free PDF
Free: the 2026 Hospital Training Catalog
Six AI-simulation courses your clinical staff actually finish — ready to drop into next year’s plan.
GET STARTED
See what the next generation of healthcare training looks like.
Whether you run a faculty, a residency program, a clinical training department, or a continuing education operation — a 20-minute demo, tailored to your context, is the fastest way to know if this fits.
Thanks — we’ve got your request.
We’ll be in touch within one business day.