AI Accountability Coach - Multi-Channel Coaching System

Behavioral change is hard, and most coaching tools are either too passive (push notifications) or too expensive (human coaches). This client needed an AI-powered accountability coach that could engage users in genuinely personalized, context-aware coaching conversations across WhatsApp, SMS, and email - at scale.

We built a multi-agent system that adapts its coaching style and communication frequency to each user's goals, progress patterns, and stated preferences.

• **LangGraph multi-agent architecture**: A coordinator agent routes between specialized agents - a goal-setting agent, a progress-assessment agent, a motivational coach agent, and a scheduling agent - each with access to MongoDB user state
• **Make.com automation layer**: Webhook-triggered Make.com scenarios handle channel routing, message formatting, and delivery scheduling without custom infrastructure
• **Twilio integration**: SMS and WhatsApp delivery via Twilio with bi-directional conversation handling - users can respond and the system maintains session context
• **Adaptive scheduling**: The scheduling agent analyzes response timestamps and goal check-in patterns to dynamically adjust nudge frequency per user
• **MongoDB state management**: Flexible document schema stores full coaching history, goal trees, and behavioral signals per user with efficient indexing
• **GCP Cloud Run deployment**: Fully containerized, auto-scaling to zero when idle, with Cloud Scheduler triggering batch nudge cycles