AI is already treating mental health. Now what?

The state of mental healthcare and AI

More people than ever are turning to generative AI for mental healthcare. In fact, Harvard Business Review found that the top Gen AI use case in 2025 and 2026 was therapy/companionship, up from #2 in 2024.

Perhaps it’s no surprise that many are turning to AI for therapy. In the United States, over 169 million people live in federally designated Mental Health Professional Shortage Areas, with Black, Indigenous, and people of color communities facing some of the most significant gaps in care. Many people who need care cannot access it, and AI can feel like an accessible way to find support when traditional care isn’t readily available.

Current solutions are failing employers and employees alike

Traditional mental health solutions like EAPs rarely work well for employers or employees. Healthcare costs are consistently rising year over year, and the root cause of these rising costs is an outdated, one-size-fits-all clinical model. While 46% of employees have no diagnosable mental illness, 36% have mild or moderate mental illness, and 18% have serious, persistent mental illness, 100% are routed into traditional one-on-one therapy. 

Employees with different levels of need require different levels of support, yet the dominant model often treats them the same. Despite record investment from employers, 57% of workers remain dissatisfied with their mental health benefit. The rise of generative AI and modern behavioral health technologies has the potential to change this, but only if they are built to guide people to the right level of care.

Our solution: Sona by Pelago

Behavioral healthcare works best when support matches each person’s needs. That’s the principle behind Sona, a voice-first AI mental health specialist that helps people access the right care at the right time.

Sona helps people before challenges become crises. Rather than offering generic conversations or static programs, Sona builds an understanding of each member over time, remembers previous interactions, and adapts support based on changing needs.

From the very first conversation, Sona works with members to create a personalized plan instead of relying on pre-built templates or decision trees. It continuously assesses changes in symptoms, goals, and risk, helping people access the right level of support at the right time. When human care is needed or requested, Pelago clinicians remain in the loop and can step in seamlessly.

A clinical architecture for AI-guided care

Sona was designed specifically for mental healthcare, unlike general-purpose AI.

It serves as an AI front door that continuously assesses acuity, context, and comorbidities to help guide each member’s care journey. Behind the scenes, Sona runs on Pelago’s Agentic Clinical Intelligence, a system of specialized AI agents trained for specific clinical tasks such as behavioral analysis, evidence retrieval, care navigation, and member support.

For members who need higher-acuity care, Sona helps match them with the most appropriate in-network clinician based on their clinical needs, preferences, availability, and scheduling constraints. Human clinicians oversee escalations, supervise care decisions, and remain responsible for clinical judgment, ensuring members receive the right care while maintaining safety throughout the experience.

Safe by design

With AI use rising and the mental health provider shortage growing, ChatGPT may now be one of the largest sources of mental health support in the country. The problem is that generic chatbots are unsafe and unsuitable mental healthcare providers. They were initially trained to inform users, not influence them, challenge them, or guide them through care.

The question isn’t whether AI will play a role in mental healthcare, but whether the technology playing a role is built with the clinical safeguards, human oversight, and care pathways needed to support people safely. Sona was built with safety at the forefront, and shows promising safety data in early pilot outcomes. It continuously screens every conversation for signals of elevated risk, including indirect and ambiguous language, the way risk often presents in the real world. Pelago clinicians supervise Sona’s operation, reviewing performance, refining protocols, and retaining authority over every clinical decision. 

The future of mental healthcare

People are already turning to AI for mental health support, whether the healthcare system is ready for it or not. Employers have an opportunity to offer a solution designed specifically for behavioral healthcare, one that expands access, personalizes support, and connects people to clinicians whenever human care is needed.

Sona gives employees an immediate place to start, while helping ensure every person receives the level of care that’s right for them.