Redefining Healthcare: Driving Results with Extractive Summarization and Eligibility BOT Implementation
Client Overview
The client is a pioneering healthcare enablement organization headquartered in the United States, focusing on transforming the healthcare experience for members, payers, and providers. The client offers end-to-end health plan management services, combining technology and domain expertise to drive better health outcomes, reduce administrative burden, and improve operational efficiency. With a growing portfolio of national and regional health plans, they are committed to delivering high-quality service on a scale.
Reimagining Member Experience Through Intelligent Automation
Customer Service Representatives (CSRs) were spending an average of 10–12 minutes per call due to manual navigation through over 1,000 benefit documents.
Additionally, summarization notes fed into the CRM lacked consistency and critical insights, negatively impacting member satisfaction and resolution times.
To streamline operations and elevate service delivery, the client sought an AI-powered solution to reduce Average Call Handling Time (AHT), enhance call summary quality, and improve response accuracy through automated benefit data extraction.
Client Requirements:
Reduce Average Call Handling Time (AHT):
Bring AHT down from 10–12 minutes to 6–8 minutes to boost member satisfaction and reduce operational strain.
Enhance Call Summarization Quality
Ensure consistent, accurate, and actionable summaries reflecting each call's core outcomes and member needs.
Automate Benefit Information Retrieval
Eliminate manual document searches by CSRs and enable instant access to relevant plan details from over 1,000 Summary Plan Documents (SPDs).
Accelerating Documentation Intelligence with AI Automation
Indium partnered with the client to deploy an AI-first solution tailored specifically for healthcare operations, combining extractive summarization, contextual data extraction, and document intelligence.
Extractive Summarization of Call Transcripts
Leveraged OpenAI's GPT-3.5 Davinci model to generate intelligent, structured summaries of customer call transcripts. We built a Minimum Viable Product (MVP) to fine-tune model outputs for business-relevant prompts, significantly improving the quality of notes ingested into the CRM platform.
Smart Eligibility BOT with Contextual Extraction
Utilized fine-tuned GPT models to extract benefits grid data from SPDs. Employed a document pipeline with overlapping chunking and context injection strategies to handle large, complex documents and accurately surface eligibility data in real-time.
Optimized Cost-to-Performance Strategy
Designed a cost-effective model for summarization and extraction workflows, considering token consumption, audio length, and inference quality to ensure scalability across high call volumes.
From Strategy to Impact:Tangible Outcomes in Engineering Productivity and Accuracy
01
9–10 Minute Reduction in Average Call Handling Time (AHT)
Streamlined information access and intelligent summaries reduced call durations, helping CSRs focus more on member empathy and faster resolution.
02
Improved Summarization Quality with Key Outcomes Highlighted:
Summaries now consistently included accurate, relevant details, improving post-call actions and enabling better analytics.
03
99% Accuracy in 90% of Member Queries
The eligibility BOT achieved near-perfect accuracy for most benefit-related questions, empowering CSRs with instant answers and driving first-call resolutions.