BalansBuddy
AI-Powered Debt Management Platform
The Problem
People in debt had to spend 50+ hours on paperwork per application. Existing solutions were chatbots that answered questions but didn't do the actual work. The rejection rate stood at 55% — more than half of all applications failed, often due to incomplete or incorrect documentation.
- 50+ hours of administrative work per application
- 55% rejection rate on debt relief applications
- Chatbots that answered questions but didn't do the work
- Counselors overwhelmed with paperwork instead of helping people
- No multi-language support for diverse populations
The Solution
We built BalansBuddy: a multi-portal platform (client, counselor, municipality) with Ember AI as an autonomous agent. Not a chatbot — an agent that handles creditor correspondence, analyzes bank statements, and detects legal escalations.
Client Portal
Guided intake, document upload, real-time status tracking
Counselor Portal
Case management, AI-assisted review, bulk processing
Municipality Portal
Oversight dashboard, compliance reporting, budget tracking
Ember AI Agent
Autonomous creditor correspondence, bank analysis, legal detection
Ember AI doesn't just respond to questions. It proactively handles creditor emails (<5 min response, 24/7), analyzes bank statements to detect patterns, recognizes 50+ creditors automatically, and flags legal escalations before they become problems. All in 6 languages.
Results
The impact was immediate and measurable:
- 98% time reduction: Application time dropped from 50 hours to 1 hour
- Rejection rate 55% to 5%: Complete, accurate applications through AI assistance
- Creditor response <5 min: 24/7 automated correspondence handling
- Counselor capacity doubled: Focus shifted from paperwork to people
- 6 languages supported: Serving diverse communities without language barriers
- 50+ creditors recognized: Automatic detection and appropriate handling
Technical Architecture
Built for security, scale, and reliability:
React + TypeScript
Type-safe frontend across all three portals
Node.js + PostgreSQL
Robust backend with full audit trail
OpenAI GPT-4
Powering Ember AI's autonomous capabilities
Redis + BullMQ
Reliable async job processing for creditor correspondence
AES-256-GCM
End-to-end encryption for all sensitive financial data
Key Insights
Building BalansBuddy taught us critical lessons about AI in social services:
- Agent vs. chatbot: The difference between answering questions and doing the work is everything
- Trust through transparency: Every AI action is logged and reviewable by counselors
- Multi-portal complexity: Three user types with completely different needs, one coherent system
- Encryption is non-negotiable: Financial data requires military-grade security
- Language inclusivity: Debt doesn't discriminate — neither should the solution
Relevance for Clients
This project demonstrates our ability to:
- Build AI agents that autonomously handle complex workflows
- Design multi-portal platforms with role-based access
- Handle sensitive data with enterprise-grade security
- Deliver measurable impact: 98% time reduction, 10x fewer rejections
If you need an AI solution that doesn't just answer questions but actually does the work — we've built it.
Timeline
Domain analysis: 50+ interviews with counselors, clients, and municipalities. Identified the core problem: chatbots don't do work.
Multi-portal design. Ember AI agent architecture. AES-256-GCM encryption layer. 6-language NLP pipeline.
Three portals built in parallel. Ember AI trained on creditor patterns. 50+ creditor recognition system. BullMQ job queue for async processing.
Live deployment. 98% time reduction confirmed. Rejection rate dropped from 55% to 5%. Counselor capacity doubled.