Saara Huhanantti is Director of Product and co-founder of Project 5/5 at Otherkind, where a multidisciplinary team builds AI products for youth mental health and crisis support. Her expertise lies in low-threshold youth mental health work, crisis counseling, and non-profit leadership, grounding her current focus on how AI can be introduced responsibly into sensitive care contexts. In her work, safety and ethics are treated as core design principles, ensuring that AI is developed to reinforce human care rather than replace it.
Saara Huhanantti
Speaker
Designed Beyond Compliance: Complementing Human Support, Not Replacing It
Speech
Sekasin chat is Finland's most widely used low-threshold mental health service for 12–29-year-olds, offering free, anonymous one-to-one support from professional counselors and trained volunteers. Despite this reach, only 26% of the 160,000 yearly contact attempts made it through the queue to a conversation (2024 statistics), revealing a persistent capacity gap between demand and available human support.
In response, AI Mood Support, a 24/7 low-threshold mental health chatbot, was developed in collaboration with Sekasin and launched in January 2025. Rather than replacing human care, the chatbot was designed to fill gaps in the existing system, offering a "no-threshold" option during off-hours and peak times, and functioning as a stepping stone toward further help.
Development was guided by a question broader than what regulation requires: what should AI actually do in this context, and what genuinely serves the young person. Meeting GDPR, EU AI Act, and data protection requirements formed only the baseline. The real work of development centered on safety and ethics: active suicidality moderation, clinically validated content, no profiling or tracking, a logic of care rather than engagement, a clear human-AI distinction, and safeguards against attachment formation. Messages indicating suicidal ideation or self-harm are never handled by AI. They are continuously monitored and routed to dedicated, human-built flows developed with the Suicide Prevention Center.
In the first 18 months, the chatbot logged 45,000 conversations, of which 25% occurred outside human-service opening hours, directly addressing the access gap that motivated the project. A separate, privacy-first analysis of 22,600 conversations and 331,000 messages examined what young people brought to the chatbot and how they responded to it. Topic analysis showed their primary need was to be heard, far more than to receive information, yet 75% of survey respondents reported positive impact, most often citing tips and advice, improved mood, and new perspectives. Sentiment analysis showed an average 11% rise in mood across conversations, with improvement in 9 of 10 topic categories, most pronounced among users who arrived most distressed. One user reflected: "For me it was even easier to talk to the bot than to a human."
The same architecture, including its guardrails, validated content, and human-first design, has since been adapted for other populations, such as victims of domestic violence, parents, and sport coaches. A companion Volunteer Training Tool has also been developed, using AI to simulate client conversations so volunteers can practice safely and receive structured feedback.
These findings suggest that carefully bounded, ethics-first AI tools can extend access to youth mental health support without displacing human connection, offering a model for how technology and frontline care can work together responsibly.