Automating Healthcare Fax Processing for Janus Health

Thousands of Faxes, Zero Standardization

Janus Health is an RCM SaaS provider that processes incoming faxes containing critical patient and billing information for healthcare organizations. The scale and complexity created compounding problems:

• Data Complexity: Faxes contained intricate medical and billing information, often handwritten, filled with industry-specific jargon. Training staff to accurately interpret this data was difficult and slow.
• Volume: Thousands of faxes received daily, creating a massive workload that needed to be processed quickly to maintain operational efficiency.
• Unstructured Format: Faxes varied widely in structure and format — no two looked the same — making consistent data interpretation challenging for both humans and automated systems.
• Long Tail of Values: The variety of data types and handwriting styles made it nearly impossible to hard-code solutions, ruling out traditional automation approaches.

The manual nature of processing these faxes led to high labor costs, delays, and a higher risk of errors — ultimately affecting Janus's billing cycles and their customers' satisfaction.

A Hybrid AI Approach — Not AI for Everything

We built a custom AI-powered system designed to automate the extraction of key information from faxes, significantly reducing human intervention. Our approach combined traditional deterministic code with targeted machine learning — our core philosophy in action:

• Hybrid AI-Driven Architecture: Classic non-AI code handled straightforward data extraction. Deep learning models, OCR, and computer vision were integrated specifically to interpret handwritten and unstructured text — the tasks where AI genuinely outperforms rule-based alternatives.
• Iterative Validation Process: The system automatically validated outputs and refined results when errors were detected. This iterative data pipeline ensured high accuracy before final human review.
• Targeted AI Use: Multiple AI models deployed strategically for specific tasks — handwriting recognition, NLP for text interpretation, OCR for document processing — rather than throwing a single LLM at the entire problem.

From Analysis to 80% Automation in 3 Weeks

• Data Analysis: Analyzed sample faxes and worked closely with the Janus team to understand processes and extraction requirements.
• Iterative Development: Continuously refined the system based on client feedback, addressing challenges like data variability and ambiguous handwriting.
• Rapid Results: After just 3 weeks of development, we automated 80% of the faxes, extracting 100% of the required data for those documents. The system evolved to 99% automation as it matured.

$4 Million in Cost Savings

• 99% Automation: Automated 99% of data extraction, dramatically reducing human labor, processing times, and errors.
• $4 Million Saved: Substantially reduced reliance on manual labor, delivering $4 million in cost savings.
• Faster Billing Cycles: Accelerated Janus's billing cycles, improving cash flow for their health system customers.
• Improved Accuracy: The hybrid approach led to fewer errors, ensuring compliance with industry standards.
• Competitive Edge: Speed and reliability improvements gave Janus a distinct advantage over competitors in the healthcare space.