shamiriAI

AI-native tiered mental healthcare.

Making providers better, therapies more potent, and care more accessible — AI embedded across every tier of the stepped-care system.

The challenge

Two structural problems

The treatment gap persists because of fundamental constraints — not lack of effort.

Not enough providers

Fewer than 10 mental health professionals per 100,000 people globally — and fewer than 1 per 100,000 in most low-resource settings. The result is long wait times and care systems that reach only a fraction of those in need.

Treatments are ineffective

Meta-analyses spanning six decades show that psychotherapies for youth depression outperform no treatment by roughly 12 percentage points — a modest effect that hasn’t meaningfully improved since the 1960s.

What AI enables

Three capabilities

AI can overcome structural barriers and be a force for more effective, more accessible care.

Providers

Making providers better

AI-augmented supervision gives every lay provider access to structured, timely feedback — closing the gap between training and practice without requiring more licensed professionals.

Therapies

Making therapies more potent

Network analysis and session-level data reveal which intervention components drive outcomes — enabling iterative improvement of treatment protocols based on evidence, not intuition.

Reach

Expanding reach

AI-guided self-care tools extend structured mental health support beyond the clinic — making evidence-based interventions available to anyone with a phone, with clear escalation pathways when human care is needed.

The system

AI-native stepped care

A four-tier system with AI embedded across each level — not replacing humans, but making every tier more effective.

Tier 1: Licensed Professionals

Psychologists and psychiatrists manage complex and high-risk cases. AI provides decision support — surfacing relevant history, flagging risk patterns, and suggesting evidence-based protocols.

Tier 2: Semi-Professionals

Supervisors recruit, train, and oversee lay providers. AI augments their work — automating session review, generating fidelity scores, and focusing supervision time where it matters most.

Tier 3: Non-Professionals

Lay providers deliver group-based interventions in schools and communities. AI provides real-time feedback on session quality, identifies skill gaps, and personalizes coaching recommendations.

Tier 4: AI-Guided Self-Care

Structured, validated self-guided interventions for young people — embedded within the stepped-care system with clear escalation pathways to human providers when needed.

Licensed professionals

AI-informed decision support

Semi-professionals

AI-augmented supervision & training

Non-professionals

AI-driven feedback on group interventions

AI-guided self-care

Structured interventions with escalation pathways

Where we are today

We've built and piloted the first component — AI-augmented supervision for lay providers.

What we've built

An AI system that processes raw therapy audio in English, Swahili, and Sheng — transcribing sessions, extracting prosodic features, and generating structured fidelity feedback for supervisors. Weeks of manual review compressed into hours.

What we've shown

A non-inferiority trial demonstrating that AI-generated supervision feedback is comparable to expert human supervision across key fidelity dimensions — with dramatically lower cost and latency.

Built on evidence

5

RCTs

11,500+

youth

100,000+

in delivery

50+

publications

JAMA PsychiatryThe LancetWorld PsychiatryNatureBMJ