
In behavioral healthcare, the shortage of therapists can have serious consequences for people who are ready to seek help but cannot find it when they need it. Demand for mental health care has become increasingly visible, but access can still depend on finding an available clinician, waiting for an appointment and arranging life around a scheduled session. For someone struggling with depression, anxiety or another behavioral health condition, those barriers can determine whether treatment begins at all. Synchronous telehealth has made it easier to connect patients and clinicians remotely, but it still depends on one clinician being available for each patient at a particular time.
I encountered that capacity problem from several directions before I began exploring how technology could extend access to psychotherapy without requiring a clinician to be present for every therapeutic interaction. While working on behavioral health strategy at a major healthcare organization, I saw the effects of wait times, network limitations and referrals that never resulted in a member receiving care. At the individual level, scheduling, cost, stigma and discomfort with beginning therapy through a live conversation could keep someone from taking the next step. I began to consider whether an asynchronous model could preserve the therapeutic value of psychotherapy while reducing the clinician time required to deliver care.
Designing an asynchronous psychotherapy platform required decisions about which elements of care could function without a live clinician, where technology could reduce dependence on scarce clinician time and where clinical expertise needed to remain in human hands. Over several years of deployments with health plans and provider organizations, we developed an architecture that uses recorded psychotherapy as the clinical foundation, with technology supporting matching, sequencing, engagement and measurement and human oversight incorporated through clinical review, coaching and escalation.
Keeping Clinical Expertise at the Center of Asynchronous Care
Group therapy offered a useful clinical precedent as I considered what could be delivered asynchronously. An individual patient does not have to be speaking at every moment to benefit from a therapeutic encounter. A patient can recognize aspects of their own experience while hearing someone with a similar problem work through it with a skilled clinician, and observe therapeutic techniques in practice and begin to understand how change develops over time.
We applied that principle by building a library of de-identified recorded psychotherapy sessions between licensed therapists and patients. Another clinician reviews each session, and the sessions are organized into series that allow members to follow a client’s progression over time.
Prerecorded psychotherapy gave us a way to extend the reach of a clinician’s work while keeping therapeutic content under clinical control. Every session is delivered by a licensed professional and reviewed by another clinician before it becomes available to members. We kept generative AI out of that process because allowing software to create or alter the clinician’s words would remove that level of review.
With recorded sessions, we know what therapeutic content a member will receive and when they will receive it, and another clinician can review the entire session in advance. Generative systems introduce a different clinical risk because responses can vary from one interaction to the next, including responses no clinician has reviewed.
Deciding Where Technology Belongs in the Care Model
Moving psychotherapy outside a scheduled live encounter required us to account for what normally happens around that encounter. A therapist can respond to what a patient says, recommend what to work on next and notice when someone is struggling or disengaging. An asynchronous model needs technology to support some of that continuity without making clinical decisions it is not equipped to make.
We built the technology layer to assess a member’s needs, match the person with appropriate sessions and determine how those sessions should be sequenced. The system also tracks engagement and clinical outcomes so that changes in progress can signal when additional support may be needed, helping maintain continuity between assessment, therapeutic content and support as a member’s needs change.
Our matching process began with a rules-based system using a small number of clinical variables. As we accumulated data on engagement and progress, we developed more sophisticated approaches to matching and personalization, including machine-learning models that consider a broader set of information when recommending what a member should receive next. We evaluate potential uses of AI by considering what could happen if the system gets something wrong.
The consequences vary considerably depending on the role technology is performing. An inappropriate content recommendation can be identified and corrected with limited clinical impact, while an inaccurate or harmful therapeutic response may directly affect someone experiencing a mental health concern. We consider the severity and reversibility of potential errors when determining which functions can safely be automated and which require clinical oversight.
Asynchronous Care Still Needs a Human Safety System
A scheduled therapy appointment gives a clinician repeated opportunities to observe a patient and recognize when the treatment plan may no longer be appropriate. An asynchronous model has to account for those same clinical responsibilities, beginning with whether the person entering the program is appropriate for this form of care and continuing with changes that may require a different level of support.
We built validated screening and exclusion criteria into intake to identify acute or unstable presentations. Suicidal ideation, psychosis or signs of increasing acuity require human intervention, with established pathways into live or higher-acuity care under clinical oversight. Technology can help identify signals that warrant attention, but responding to a meaningful change in clinical risk requires professional judgment.
Coaches provide another layer of human support by helping members navigate the program, stay engaged and work through practical barriers that might interrupt care. Their role is clearly defined because navigation and engagement carry different clinical responsibilities from psychotherapy. A member whose needs move beyond the coach’s role must have a clear path to appropriate clinical support. Technology can help identify and coordinate those transitions, but clinical ownership remains essential when a member’s needs change.
Design the Pathway for the Person Who Chose It
We learned an important lesson by looking closely at how members moved through an asynchronous program. People had chosen an app-based, relatively private way to begin behavioral health care, often because they could engage on their own time and with less pressure to disclose their concerns immediately. But in one deployment, when the next step required members to speak with a clinician by phone, participation dropped sharply. From the member’s perspective, the phone call reintroduced an interaction some had been trying to avoid.
The fix was to stop routing members back into a modality they had already declined. The drop in participation taught us to pay closer attention to why someone chooses an asynchronous option and carry that preference through subsequent stages of care. A digital entry point provides limited value if the next step reintroduces the scheduling, disclosure or interaction barriers that kept someone from seeking care earlier.
Integrating Asynchronous Psychotherapy Into the Broader Care System
Deployments with health plans and provider organizations showed us how much an asynchronous model depends on connections to the rest of the care system. Plans and providers need to know whether a member is engaging and improving, how changes in clinical risk will be escalated and whether relevant outcomes can follow the member into subsequent care. Our platform supports those connections through structured measurement, documented escalation protocols and the return of relevant clinical information to organizations involved in a member’s care.
Some people need live, bidirectional therapy because of their condition, acuity or need for an ongoing therapeutic relationship and accountability. Others may be clinically appropriate for asynchronous care but have struggled to receive enough therapy through a model dependent on appointment availability, scheduling and willingness to begin with a live interaction. Expanding capacity requires giving that second group another credible way to receive care without treating asynchronous psychotherapy as a universal replacement for live therapy.
Several years of building and deploying this model have made me cautious about framing the behavioral health capacity problem as a choice between adding clinicians and automating their work. Clinician supply remains constrained, while transferring therapeutic responsibility to software raises clinical and accountability questions that the technology is not yet equipped to resolve. The better opportunity is to use technology selectively, extending the reach of clinical expertise by improving how people enter care, how appropriate therapy is matched and sequenced, how progress is measured and how changes in need are recognized. Expanding behavioral health capacity depends on designing care so that limited clinical resources can reach more of the people who need them.
About Sushant Gupta
Sushant Gupta is the founder and CEO of Meomind, a digital behavioral health company providing on-demand access to recorded psychotherapy supported by AI-driven personalization, care navigation and certified mental health coaches. Under his leadership, Meomind has grown to approximately 150,000 users and works with health plans, providers and employers. Previously, Gupta served as Director of Growth & Business Transformation at UnitedHealth Group/Optum, and held a senior strategy and product role at Philips focused on healthcare technology and AI-based clinical decision support. He holds an MBA from Harvard Business School and B.Tech and M.Tech degrees in Computer Science and Engineering from IIT Kharagpur.

