What AI With “Human-in-the-Loop” Means at Lyra

June 11, 2026

In mental health AI, "human-in-the-loop" is a widely used term for AI systems that include some human involvement. But the depth of that involvement, and its impact on safety, quality, and member experience, can vary dramatically.

AI has enormous potential in behavioral health. It can expand access, help people get support faster, and identify and manage risk at scale, while making care’s impact stronger and longer lasting.

The opportunity is significant, but realizing it isn’t straightforward. Mental health conversations often involve nuance, vulnerability, trauma, and risk. Supporting people effectively requires more than engaging interactions—it requires clinical expertise to ensure people receive the most appropriate support for their needs, ongoing oversight, and a strong foundation in evidence-based care.

At Lyra, we believe AI and clinical expertise are most powerful when they work together. AI can help scale support and make it available 24/7, and clinical expertise helps ensure that support remains safe, effective, and grounded in evidence-based principles.

This is what “human-in-the-loop” means at Lyra. AI helps enhance mental health support, while clinical experts and providers play a central role in how the technology is developed, tested, monitored, and used.

We apply this approach across three core areas:

1. Clinical expertise in AI development

Clinical experts help define the standards, guardrails, and evaluation criteria that shape how Lyra AI is built, tested, and improved over time.

• What our people do: Lyra’s clinical and research teams work alongside our AI and technology teams at every stage of development. They help design risk detection systems, create and validate training datasets, define safety and escalation protocols, and continuously review real-world performance to help ensure AI remains safe and effective.
• What our AI does: We’ve built our own AI technology that flags potential risk, misuse, or situations where a different level of support may be needed, helping route those cases to the appropriate clinical experts and in-the-moment human support.

2. Provider-led care and AI review

Effective mental health care is built on strong therapeutic relationships and evidence-based support. AI can help strengthen care availability and reduce administrative burden, giving providers more time to focus on their clients.

• What our people do: Even as we integrate AI into the care experience, our care model remains firmly provider-led. Clinicians have full autonomy over care plans and treatment decisions. Providers also review client AI interactions as part of ongoing quality oversight and use those insights to better understand client needs, tailor support, and help people make progress more quickly.
• What our AI does: Our proprietary summarization tools review key themes from AI-client interactions, making it easier to understand how people are using the technology, identify patterns that may warrant attention, and evaluate how well the AI is supporting members. This ongoing visibility helps providers maintain clinical oversight and inform continuous improvement.

3. Crisis detection and expert intervention

When elevated risk is identified, human experts step in.

• What our people do: Our 24/7 care team is trained to support members during high-risk situations involving suicidal ideation, self-harm, or homicidal and violent ideation. When risk is identified, the team follows Lyra’s intervention and outreach protocols to de-escalate risk, manage crises in the moment, and connect individuals with appropriate follow-up support.
• What our AI does: Our AI model helps detect signs of potential risk, whether explicitly stated or implied, and immediately routes those interactions to our care team for clinical review and intervention.

Clinical innovation is essential to AI innovation

As AI becomes more common in mental health care, technical innovation alone isn't enough. Organizations also need safeguards, oversight, and clinical expertise to help route people to the right support at the right time and ensure they receive the most effective care for their needs.

For us, human-in-the-loop is more than a layer of oversight. It's a commitment to embed clinical expertise, evidence-based care, and ongoing evaluation into every stage of how mental health AI is developed, tested, and delivered.