Milo App.

The problem

Therapy patients in active recurring sessions face a recurring problem: the insights and emotions from one week get forgotten before the next session. Therapists start cold, the patient struggles to remember what happened, the first 10–15 minutes of every session is reconstruction rather than progress, and the patient can't see their own progress across weeks because nothing systematic captures it.

The downstream effects are real: sessions feel repetitive (same recovery from cold start each week), patterns and behavior loops stay invisible to both patient and therapist, and despite paying $150–300/session, patients often leave therapy without the cumulative insight that consistent reflection would produce.

The solution

Milo solves the between-session gap with three connected AI features: a micro-journal that captures emotion, trigger, and theme in 5 minutes via voice or text; an Emotion Pattern Dashboard that detects recurring themes and behavior loops across weeks; and a one-tap Pre-Session Summary that generates a structured brief before the next therapy session.

The framing is precise: not a chatbot replacing therapy, not a generic journal app, but a cognitive clarity tool that makes the user's existing therapy more effective. The patient walks into the next session with a structured summary in hand: key emotional events, emerging patterns, suggested questions — and the therapist gets to skip the cold-start reconstruction phase.

How it works

Flow: user opens Milo → micro-journal prompt (calendar-aware, personalized to focus area) → 5-minute voice or text entry → AI structures the entry with emotion (explicit + inferred), trigger, theme → entry feeds the pattern dashboard → before next therapy session, one tap generates Pre-Session Summary.

Session length is inferred from the first message: "just a quick check-in" triggers short mode (3 follow-ups); otherwise long mode (7 follow-ups) applies. Edge cases handled: short/disengaged answers ("fine," "idk") get one gentle follow-up, then respect the disengagement; 3+ short answers in a row → skip to summary early.

Primary model: Claude Haiku 4.5, selected via evaluation: 6 models × 5 criteria (accuracy, tone, guardrails, question quality, summary quality). Claude Haiku 4.5 scored highest (9/10) — held the 3-question limit, named specific underlying patterns in summaries, maintained therapeutic-adjacent tone safely. Required inputs: User Message (free text), Emotional State (explicit + inferred), Intent.

Who it's for

Primary persona — Laura, 34: urban resident in a large city, Marketing Manager at a tech company, Master's degree, above-average income. Quote: "I want to understand myself better and take care of my emotional health." Already in recurring therapy, already invested in wellness (podcasts, gym, premium apps), high workload, limited time for reflection.

Primary segment: therapy patients aged 25–39 — urban professionals already paying for recurring sessions, self-purchase the subscription, no therapist or institutional gatekeeper required. Secondary: people with anxiety or stress management who don't have therapy sessions but want structured between-event reflection.

Why it matters

1B+ people globally have mental health conditions and 54% never get professional help. The mental health workforce shortage is structural — 13 mental health workers per 100,000 population globally as the median. The AI-enabled mental health segment is at ~23% CAGR, driven by rising demand and accelerating AI acceptance among young adults.

Milo's structural bet is on the between-session wedge: rather than competing with therapists (who resist AI replacement) or building a generic mental-health chatbot (regulatory minefield + low trust), the product augments existing therapy, builds the patient's own pattern recognition over time, and creates shareable insight at the therapist boundary (the Pre-Session Summary is the artefact). Hallucination risk is bounded by scope: the product doesn't diagnose, prescribe, or replace clinical judgment — it captures and structures the user's own reflection.

At a glance

Project

Milo App

Built by

Maidah Mir, Miglė Jaraitė, Mijeong Kim & Nwando Nwafor

One-liner

AI between-session wellness companion for therapy patients — voice/text micro-journaling, emotional pattern detection, and one-tap therapy session prep.