The Problem

Home health nurses and aides spend a huge share of every workday on documentation—OASIS assessments, visit notes, care plan updates—usually typed from memory after leaving the patient's home. This drives burnout and turnover (60%+ annually in home health aides) and creates compliance risk when notes are incomplete or inconsistent with what Medicare requires.

The Solution

CareScribe is an AI scribe built specifically for home health visits. A clinician turns it on at the start of a visit (with patient consent); it listens to the natural conversation and care activity, then auto-generates a structured, agency-compliant visit note—mapped to the specific fields the clinician's EHR and payer require (OASIS items, skilled nursing note, POC updates)—ready for a 60-second review and one-tap submission.

How It Works

• Clinician starts a visit session on their phone/tablet before entering the home
• On-device/cloud speech-to-text captures the conversation and any verbal vitals/observations
• An LLM trained on home health documentation templates extracts and structures the note: vitals, ADLs, skilled interventions performed, patient response, any red-flag findings
• The app cross-checks the draft against the agency's required fields and flags anything missing ("Wound measurement not mentioned—confirm?")
• Clinician reviews, edits, and exports directly into the agency's EHR (via copy/paste, PDF, or API integration with major home health EHRs)

Why Now

Speech-to-text and LLM summarization are finally accurate and cheap enough to run this reliably in noisy home environments, and home health agencies are under increasing financial pressure (value-based purchasing, tighter audits) making documentation efficiency and accuracy a budget-holder priority, not a nice-to-have.

Business Model

Per-clinician monthly subscription sold directly to home health agencies (not individual consumers), priced against the cost of clinician turnover and overtime hours spent on notes—easy ROI story for agency administrators.

MVP

• Mobile app (iOS/Android) for individual clinicians: start/stop visit recording, works offline-first with sync
• Speech-to-text pipeline (Whisper or similar) tuned for home environments (background noise, multiple speakers)
• LLM-based note generator supporting 2–3 of the most common documentation formats (standard visit note + one OASIS-style template) for a pilot agency
• Simple review/edit screen showing the draft note with flagged missing fields
• Manual export (copy/paste or PDF) — no deep EHR integration required for v1
• Pilot with a single home health agency (10–20 clinicians) to validate time savings and note accuracy before building EHR integrations

Core Concept & Value Proposition

SoleAgent is a plug-and-play WhatsApp and SMS AI agent that independent shoe boutiques deploy to their customer base in minutes. Instead of competing with Zappos on algorithm budgets, a small shoe shop can now offer every customer a personal "shoe advisor" who texts them, learns their fit profile, recommends from live inventory, and follows up when something new arrives that matches their taste. The result: fewer returns, higher repeat purchase rates, and a direct owned channel with every customer — no app required, no developer needed.

The Problem It Solves

Independent shoe retailers are getting crushed by three compounding problems:

• Crippling return rates (25–40%) caused by online sizing mismatches that destroy margins
• Zero post-purchase relationship — customers buy once and drift to Amazon or Nike.com
• No affordable personalization tooling — enterprise solutions like Dynamic Yield cost $50K+/year, completely out of reach for a boutique

How It Works (User Flow)

• Boutique onboarding (10 minutes): Owner signs up, connects their inventory via a Google Sheet, Shopify CSV export, or a direct Shopify integration. They add a WhatsApp Business number provisioned through SoleAgent.
• Customer opt-in: Boutique texts their existing customer list a simple "Hey, meet your personal shoe advisor — reply START." QR code on receipts captures walk-in customers.
• AI fit profile capture: The agent opens a friendly 4-message conversation — current size, preferred brands, foot width ("do shoes usually feel tight?"), and primary use (everyday, athletic, dress). This is stored per customer.
• Smart recommendations: When a customer asks "what's new in a size 10 wide for walking?" the agent searches the boutique's live inventory and responds with 2–3 matches, a short description, and a direct buy link.
• Proactive reengagement: When new inventory arrives, the agent automatically texts matched customers ("Your size just came in for that New Balance style you liked — want a link?").
• Return reduction loop: Post-purchase, the agent follows up: "How's the fit?" Responses train better future recommendations and flag issues before a return is initiated.

MVP Scope (Weeks, Not Months)

The MVP is intentionally razor-thin:

• Twilio WhatsApp Business API for messaging infrastructure
• OpenAI GPT-4o with a structured system prompt + tool calls to query the inventory sheet
• Google Sheets or Airtable as the inventory backend (no custom DB needed at launch)
• A simple Next.js/Vercel onboarding form for boutique owners to connect their sheet and provision their number
• Basic customer profile store in Supabase (size, width, past brands, last purchase)
• Cron job that checks new inventory rows and triggers proactive outreach to matched profiles

Total stack: Twilio + OpenAI + Supabase + Vercel. One developer, three to four weeks to a working product.