Back to work
n8nCase study

Pediatric Sleep Study Prep Concierge AI Assistant

AI-powered parent concierge system that triages incoming messages for medical emergencies, routes non-urgent queries to an OpenAI-powered prep assistant, and applies a secondary safety check before responding ensuring safe, intelligent, and real-time support throughout the sleep study preparation process.

n8n · workflow canvasproduction build
Pediatric Sleep Study Prep Concierge AI Assistant workflow

Context

Clinical staff manually monitored and responded to parent messages before pediatric sleep studies, with no structured triage system to flag emergencies, no AI-assisted prep guidance, and no safety validation layer to prevent inappropriate responses from reaching parents.

Approach

Built two interconnected n8n sub-workflows, the main pipeline performs emergency triage on every incoming parent message, escalates urgent cases to the clinical team via Gmail and Slack, and routes non-emergency messages to an OpenAI Prep Concierge Agent with memory and tools, followed by a secondary safety check before delivering responses. A global error handler monitors all stages and alerts the Operations Manager on failures.

Architecture

  1. Webhook POST (parent message)
  2. Extract message
  3. Emergency triage check
  4. Emergency path: Gmail + Slack (Clinical Team)
  5. Gmail auto-reply (Parent)
  6. Google Sheets log
Non-emergency path
  1. AI context builder
  2. OpenAI Concierge Agent (GPT + Memory + Tools)
  3. Secondary safety check
  4. Safe: Gmail (Parent) + Google Sheets CRM log
Flagged
  1. Response withheld
  2. Global Error Trigger
  3. Google Sheets error log
  4. Gmail (Operations Manager)

Results

Under 1 minuteEmergency escalation time
~6 hrs/weekClinical staff time saved
90%Manual triage workload reduced