Mindyai builds secure B2B infrastructure with strict ethical parameters to safeguard worker agency and clinical safety.
“Mindyai does not provide clinical diagnosis, medical treatment, crisis intervention, or emergency services. Where urgent support is needed, Mindyai directs users to local emergency and crisis-support options and organisation-approved human support pathways.”
Mindyai does not claim to diagnose mental health disorders, evaluate PTSD severity index, screen for clinical depression, or forecast suicidal ideation. AI services only summarize voluntary input into objective formats to assist human care coordinators.
All AI summaries are marked as 'AI-assisted draft' and require active human verification. The system never automates casework or implements response decisions without human professional sign-off.
We prohibit managers from accessing raw employee text entries. Aggregate insights are locked (requiring cohorts larger than 10 check-ins) to prevent re-identification or automated performance review actions.
Users decide whether they want AI summary models to evaluate their free text. Consent settings can be toggled instantly, triggers an automatic purge of the check-in's voluntary text logs.
Our Bedrock summarization templates are transparently documented. Wellbeing professionals can cross-reference the AI summary directly against raw user scores to inspect reasoning.
Voluntary check-in texts are stored in encrypted states using AWS KMS. Data retention schedules default to automatic redaction, and user-initiated deletion requests are resolved immediately.
Mindyai is built around structured triage workflows rather than real-time threat calculations. If a user enters crisis keywords, the dashboard overrides standard priority rules to flag the case as URGENT in the care coordinator's queue and directs the employee immediately to telephone hotlines.
A safety lead or authorized clinician must be configured during organization onboarding before any escalation pathways can be activated. We do not enable standalone AI alert routing without verified human contacts.