Yuna AI

Yuna AI is an employee mental health coach sold to employers, positioning itself as a 'modern replacement for the EAP.' It follows an AI-first, human-backed approach: everyday conversations are handled entirely by the AI coach, and only high-risk signals escalate to the 988 crisis line or a licensed therapist. On the HR side, only de-identified aggregate data is visible; no individual conversation is ever exposed.

Core capabilities

• Talk anytime: open by voice or text with no appointment and no intake questionnaire, unlimited within the subscription, and the vendor says response is so fast there is almost no wait
• Risk routing: after detecting crisis language, it first de-escalates inside the session, then surfaces 988 and emergency resources, and matches a clinician when needed
• De-identified dashboard: the Pulse panel shows HR aggregate trends in engagement, outcomes, and topics; individual conversations never come into view
• Benefits navigation: routes employees back to benefits the employer already offers, such as an existing EAP or financial coaching
• Self-guided practice: meditation audio, a daily journal, and mood ratings before and after sessions are built into the flow
• Two entry points: the employee app and the HR admin console are separate, with permissions and data definitions kept apart

Where it fits Small and midsize employers that already have an EAP almost nobody uses and do not want to add headcount to benefits; it also suits employees on night shifts, people who cannot get an appointment, or those who want someone to practice with between therapy sessions. Brokers and PEOs can slot it into client benefits packages. It does not diagnose, does not prescribe, and does not take on crisis intervention.

Differentiation Against human-staffed models such as Spring Health and Lyra, it pushes service cost down to a per-head subscription, and the vendor says participation is far higher than with a traditional EAP. The underlying foundation model vendor is not disclosed; the real in-house work is the application logic of the memory architecture and conversation orchestration.

Editorial take Plainly: the only publicly visible reputation is a small-sample app store rating, there is almost no third-party evaluation on the enterprise side, several of the vendor's outcome figures are stated inconsistently across its own materials, and the efficacy evidence comes from a retrospective paper co-authored by its own team. The therapist network covers the United States only, the terms state outright that the AI is not a person and that there is no real-time human monitoring, and enterprise quotes require a meeting. Employers willing to accept AI-first care and to press hard on SOC 2 and HIPAA claims should give it a look; companies that need live human coverage should not replace their EAP with it yet.