WhatsCovered
Chatbot powered by LLMs trained on insurance data from South Africa's biggest insurers to help consumers understand their medical aid plans.
Project Overview
WhatsCovered is a suite of AI-powered chatbots designed to help South African consumers navigate the complexities of their medical aid plans. The chatbots are trained on comprehensive datasets from the country's major medical insurers, enabling them to provide accurate, personalized information about coverage, benefits, and claims processes.
The solution was developed to address the significant information gap that exists between medical aid providers and their members, who often struggle to understand what procedures and medications are covered under their specific plans.
Project Highlights
- •Custom LLMs trained on South African medical aid data
- •Natural language interface for complex coverage queries
- •Integration with insurer systems for real-time information
- •Multi-channel deployment (web, WhatsApp, mobile app)
- •Compliance with healthcare data protection regulations
The Challenge
The client, a health insurtech startup, faced several significant challenges in developing an effective solution:
Complex Information
Medical aid plans in South Africa are notoriously complex, with numerous exclusions, conditions, and benefit structures that vary widely between providers and plans.
Data Access
Obtaining comprehensive, structured data from medical aid providers to train the AI models required complex negotiations and data processing.
Accuracy Requirements
Given the critical nature of healthcare information, the solution needed to achieve exceptionally high accuracy rates to be trustworthy and useful.
Regulatory Compliance
The solution needed to comply with strict healthcare data protection regulations while providing personalized information.
Our Solution
MSYSTEMS developed a comprehensive AI-powered solution with several innovative components:
Specialized LLM Training
We developed a methodology for fine-tuning large language models on structured and unstructured medical aid data, creating specialized models for each major insurer's policies and benefits.
Natural Language Interface
We built a conversational interface that allows users to ask questions in natural language about their coverage, using everyday terms rather than insurance jargon.
Multi-Channel Deployment
We deployed the chatbots across multiple channels including a web portal, WhatsApp (the most popular messaging platform in South Africa), and a dedicated mobile app.
Secure Authentication
We implemented a secure authentication system that verifies user identity while maintaining compliance with data protection regulations, allowing for personalized responses.
Human Fallback System
We designed an intelligent escalation system that transfers complex queries to human agents when necessary, with the AI providing the agent with relevant to human agents when necessary, with the AI providing the agent with relevant context and information to resolve the issue efficiently.
Results & Impact
Reduction in call center volume
Query resolution accuracy
Monthly active users
Member satisfaction scores
WhatsCovered has transformed how South African consumers interact with their medical aid providers. The chatbots have significantly reduced the burden on call centers while providing members with immediate, accurate information about their coverage.
The solution has been particularly impactful in helping members understand preventative care benefits that were previously underutilized, leading to increased usage of these services and potentially better health outcomes. Several major medical aid providers have now integrated WhatsCovered into their member portals and apps.