IGOR POLYANSKY
SENIOR / LEAD PRODUCT DESIGNER

SVOY CHELOVEK

Therapist Matching in Telegram



A Telegram Mini App where AI suggests the match and a human makes it

ROLE
Founder & Product Designer

SCOPE
Product strategy · Product design · Web and mobile
svoychelovekpsy.ru
@svoy_chelovek_psy_bot



Main
Application
Selection
Messages



    PROBLEM
    • Finding the right therapist is a trust problem; directories make it worse
    • The first step is the hardest — any friction kills it
    • Raw intake text overwhelms the specialist
    • "Let AI decide" is exactly what harms in mental health



    DECISIONS
    • Telegram Mini App: zero-friction first step in a familiar space
    • Client sees only a human; AI is invisible by design
    • LLM structures the request into a brief and suggests specialist candidates — to the operator, never to the client
    • Confidence per theme, urgency flags, drill-down to the client's original words
    • Data minimization: no diagnoses or medical details ever collected



    User Flow



    TRADE-OFFS
    • Human in the loop → less throughput, more trust
    • AI hidden from the client → no "AI-powered" shine, no machine at the vulnerable moment
    • Suggestion over decision → AI never gatekeeps access to help



    OUTCOME
    • Shipped 0→1: independently built the Telegram Mini App and LLM-assisted operator workflow
    • Specialist gets a structured brief with candidates, not raw text
    • Clear human–AI boundary: model structures and suggests; human decides and explains



    INSIGHT
    In a high-stakes domain, the design question isn't what AI can do — it's where it must step back. Here it stepped fully behind the operator: the client never meets the model, and that absence is the design.
    Updated 08.26