More Than 1 Million People Use ChatGPT for Mental Health Every Week. Jimini Health Just Raised $17M to Build What Behavioral Health Systems Actually Need.

According to the investment thesis that drove Jimini Health's March 2026 funding round, more than one million people per week have conversations with ChatGPT that include indicators of suicidal planning or intent. None of those conversations are visible to a clinician. None are aligned to a care plan. None trigger a follow-up from someone who knows the patient's history. The New York-based behavioral health technology company closed a $17 million seed round on March 31, 2026, to build the clinical infrastructure it believes that category has never had.

The Gap Between Sessions Is Where Mental Health Outcomes Are Made or Missed

Mental health care has a structural continuity problem. Most patients see a therapist or counselor once per week. Some less frequently. The therapeutic work, including processing difficult events, practicing coping techniques, and monitoring symptoms, happens between those sessions, in the hours and days when no clinician is reachable.

General-purpose AI chatbots have filled part of that gap, whether behavioral health systems intended it or not. According to data from the KFF's April 2026 health policy polling, about 16 percent of Americans who used AI for health information sought mental health guidance specifically. For millions of patients, the decision to use an AI chatbot is not a technology preference. It is a rational response to a system that provides no supervised alternative between appointments.

The clinical risk of unsupervised AI in this context is not theoretical. A general-purpose language model operating outside a clinical context has no access to a patient's diagnosis, no awareness of the specific therapeutic approach their provider selected, no visibility into prior crisis episodes or medication history, and no mechanism to escalate to a human clinician when a conversation warrants it.

Sage: Every Interaction Supervised, Every Care Decision Human

Jimini Health's platform is called Sage. The system functions as an AI behavioral health assistant designed to extend clinical support into the time between therapy sessions: checking in on patients, reinforcing therapeutic homework, providing reminders, and surfacing signals that warrant clinician attention.

The architecture is built around a single clinical principle: every care decision remains with the human provider. Every interaction between a patient and Sage is visible to the supervising clinician in real time. Sage follows the specific care plan the clinician designed for that patient. It does not perform diagnostics. It does not modify care plans. It does not operate independently of the clinical team's oversight.

M13 partner Morgan Blumberg, who co-led the investment, stated that Jimini is building the clinical infrastructure this category has never had, describing the model as providing real supervision, real clinicians, and real oversight at every step. This framing reflects a deliberate positioning against both unsupervised consumer chatbots and against fully autonomous clinical AI tools that have faced regulatory and liability scrutiny.

The platform is not a consumer application. It is a B2B product targeting large, multi-site behavioral health organizations, where the clinical team is the customer and the patient is the endpoint. Deep integration into behavioral health EHR systems is the intended deployment model.

The Numbers

Jimini Health closed its $17 million seed round on March 31, 2026. Investors include M13, Town Hall Ventures, LionBird, Zetta Venture Partners, and OneMind. Total capital raised now exceeds $25 million. Among the backers is Andy Slavitt, who served as Acting Administrator of the Centers for Medicare and Medicaid Services under President Obama. Slavitt's involvement is strategically meaningful: direct experience with CMS reimbursement structure and federal healthcare policy is the kind of institutional knowledge that shapes how a behavioral health platform navigates payer integration and compliance.

The company has stated its intention to expand Sage's capabilities across comorbidities, additional care settings, and new patient engagement modalities as the platform scales.

What To Do About It

1. If you are building or procuring healthcare IT systems, make clinician visibility a baseline requirement for any patient-facing AI. The Jimini model is architecturally sound: the AI does not operate autonomously, and every patient interaction surfaces to the responsible provider. This is both a clinical safety architecture and a liability management posture.

2. Evaluate AI behavioral health tools against your EHR integration requirements before procurement. Platforms that cannot write interaction data back to existing clinical record systems create documentation gaps and compliance exposure. Third-party behavioral health AI that runs alongside your EHR rather than inside it increases risk, not care quality.

3. Build patient consent workflows that are specific to the use case. Patients using AI for support between therapy sessions need clear, plain-language disclosure about what the AI can and cannot do, who can view their conversations, and how to escalate if they are in crisis. General AI consent language does not cover this.

4. Understand how your patient population is already using general-purpose AI for mental health support. KFF data suggests widespread informal use is already happening. Health systems that know their patient population's existing AI behaviors are better positioned to offer supervised clinical alternatives before patients develop habits around unsupervised tools.

HRIM's Take

Behavioral health systems face pressure from every direction: workforce shortages, access gaps, payer constraints, and a patient population that has already started self-managing using tools with no clinical guardrails. Jimini Health's architecture is conservative by design, and that conservatism is strategically correct. It does not try to replace the clinician. It extends the clinical relationship into the between-session hours where most patients are currently on their own. For health IT teams building or evaluating behavioral health platforms, the clinician-supervised model is the right reference architecture. The question is not whether to deploy AI in behavioral health contexts. Every health system will. The question is whether the deployment is clinically supervised or not.