Editor’s note: Sabba Nazhand is the founder of Safar, a platform building AI tools for psychedelic integration practitioners.
When I was around twelve, I had a Michael Jordan poster on my wall. Jordan flying through space, with the words “Out of this world.”
That same stretch of my childhood, my mother was fighting severe depression.
Almost every day for a summer, she would come into my room, cry in front of that poster, and say, “They want me out of this world. They want me to die.”
That moment never left me. Not having the right support. Not being able to name what I was going through.
So I did what many of us do. I learned to perform. To achieve. To become someone others would want. I spent most of my life as a shell of myself, chasing love, recognition, safety, belonging. I attached myself to anyone and anything that made me feel wanted.
That foundation eventually crumbled. I lost myself. Depression set in, and thoughts of suicide circled. Death started to look like a north star. A connection to peace. A way to finally silence the noise.
I am telling you this because the conversation about AI in mental health care has become bloodless. Abstract. A debate between technologists and clinicians, fought in op-eds and panel discussions, while the people who actually need help keep waiting. I am not interested in that conversation. I am interested in this one: what would have changed for my mother, and for me, if we had the right care?
The Gap That Almost Took Us
My mother did not lack the desire to get better. She lacked access to consistent, sustained care. That was the early 1990s. For people like her, that kind of care has only gotten harder to find.
More than 122 million Americans live in federally designated Mental Health Professional Shortage Areas. Wait times for a new appointment run from three weeks to six months. Nearly half of behavioral health workers say the shortage has made them consider leaving the field. Three quarters fear lives will be lost because of it.
These are not statistics to me. They are the distance between my mother crying in front of a poster and someone being there to help her through it. The distance between a person reaching out and a person giving up.
We cannot train new therapists fast enough to close that gap. We can extend the reach of the ones we have.
What I Mean by “Beside the Therapist”
I do not mean a chatbot that pretends to be a therapist. I mean tools that work inside the clinical relationship, extending the practitioner’s capacity rather than circumventing their role.
The most underserved moment in mental health care is not the session. It is everything between sessions. A client has a breakthrough on a Tuesday. Their next appointment is in two weeks. By then the insight has faded, the emotional charge is gone, and the practitioner is scrambling through notes to catch up before the call.
Tools that support reflection, journaling, and pattern recognition between sessions can preserve what emerged in the room. Not replacing the therapist’s interpretation, but surfacing the material so the therapist has something richer to work with when the client returns.
AI documentation tools have already been shown to cut charting time by more than half. That is hours returned to actual care. Hours a burned-out practitioner with 50 clients needs.
Then there is pattern recognition across time. Humans are extraordinary at reading a room. We are less reliable at tracking subtle behavioral shifts over months. Systems that analyze mood data, sleep patterns, and engagement signals can surface trends a practitioner might miss. Not because the practitioner is careless, but because they are human, managing their own overwhelm.
A clinician at the University of Washington put it simply: therapists have historically been limited to what a patient remembers on a particular day. Now we are learning how to bring the rest of a patient’s life into treatment.
The Evidence Is Early. It Is Also Personal.
A randomized trial of 210 adults, run by psychologists and psychiatrists at Dartmouth and published in NEJM AI, tested a generative AI chatbot called Therabot. Participants with major depressive disorder saw an average 51 percent drop in symptoms over eight weeks. Participants with generalized anxiety disorder saw a 31 percent reduction. The lead researcher said the results were comparable to randomized trials of psychotherapy with human providers, in roughly half the time.
What struck me most: participants reported a therapeutic alliance with the tool approaching norms for human outpatient care. The technology did not just deliver information. It created a sense of being heard.
I know what it feels like not to be heard. To perform wellness while drowning. To need someone at 2 a.m. and have no one to call. I am not suggesting we hand everyone an app and call it therapy. I am saying continuity matters. Presence matters. And for the millions of people sitting in the gap between crisis and care, a well-designed tool is better than silence.
At Cedars-Sinai, physician-scientists are studying XAIA, a program that uses virtual reality and generative AI to deliver conversational therapy in immersive environments. They encoded more than 70 best practices from expert therapy sessions into the system. Not to replace the expert, but to extend the expert’s approach to the moments when no therapist is available. Early findings in npj Digital Medicine found the approach safe and worth further research.
Where I Come In
I work in psychedelic integration now. I build technology, including a platform called Safar, to support practitioners who guide people through some of the most profound and vulnerable experiences of their lives.
Integration is the work that happens before and after the medicine. It is where the insights from a session get translated into lasting change, or do not. The practitioner shortage that plagues mental health care at large is magnified here, where the pool of qualified, experienced guides is a fraction of the demand.
AI cannot hold space the way a skilled practitioner can. It does not know when silence is more therapeutic than words. But it can help a practitioner stay connected to 50 clients between sessions instead of 10. It can surface a journal entry from three weeks ago that holds the thread a client needs to pull next. It can flag when someone’s engagement drops off, a signal that often precedes the hardest moments, and prompt the practitioner to reach out before the crisis instead of after.
That is what responsible deployment looks like. Not a chatbot pretending to be a shaman. An instrument in the practitioner’s hand.
Name the Risks or Lose Credibility
Any honest conversation about AI in mental health has to reckon with what can go wrong.
Dependency is a legitimate concern. Research from OpenAI and the MIT Media Lab found that extended heavy use of AI chatbots can heighten loneliness and emotional reliance, the opposite of the therapeutic goal. Sycophancy is baked into many large language models, which are optimized to keep users engaged, not to challenge them.
Data privacy in mental health is a moral obligation, not a compliance checkbox. The information people share in their most vulnerable moments deserves the highest level of protection.
And there is the cultural dimension. The vast majority of counselors in the U.S. identify as white, while racial and ethnic minorities make up nearly 40 percent of the population. Tools trained on narrow datasets risk amplifying the biases already built into care. If the technology cannot serve diverse populations, it will deepen the inequities it claims to address.
None of this is a reason to reject AI in mental health care. It is the reason to build it right.
What I Wish Had Existed
I think about my mother in front of that poster. I think about the version of me that was calculating exits.
Neither of us needed an app. We needed a human being who understood what we were going through and had the capacity to stay with us through it. That human being was unavailable, overwhelmed, or did not exist in our zip code.
The mental health crisis is not waiting for a perfect solution. Millions of people are already turning to AI chatbots because they cannot find or afford a therapist. The tools are arriving whether the field is ready or not.
The question is whether we build them with clinical integrity, practitioner oversight, and real respect for the person on the other end, or whether we leave it to the market to sort out.
I know this space. Professionally and personally. I watched my mother suffer. I suffered. The system wasn’t there for either of us. That’s why I’m here.
