BrazilHealthcare and life sciences2024 – ongoingOrizon
Generative AI for medical billing rules and fraud detection
Orizon now processes 63% of medical-rule tasks automatically with generative AI
Problem
Orizon, a health tech company in the Bradesco Group, runs a platform that detects fraud in medical billing across insurers, hospitals and doctors. Its system held around 40,000 medical rules and added around 1,500 new ones each month. Business users depended on developers to query and understand the rules.
Approach
Orizon fine-tuned Llama2-code and DBRX models on its own data and business rules and served them through Databricks Model Serving. It connected a language model to Microsoft Teams so business users could ask about medical rules and get explanations without a developer. It also used the models to write code documentation.
Results
Takeaway
Putting a language model in the chat tool staff already use moved routine rule questions away from developers. The company reported freed developer time as well as a share of tasks handled automatically.
Categories
- Size
- Large enterprise
- Industry
- Healthcare and life sciences
- Business function
- Production and qualityIT and R&D
- Innovation type
- Process innovation
- AI technology
- Generative AI
- Adoption
- Off-the-shelf tool or platform
- Evidence
- Vendor case study
Sources
Source summaries
Databricks describes how Orizon used fine-tuned language models and a Teams-based assistant to handle medical rule tasks. The page reports 63% of tasks processed automatically, code documentation built in about five minutes, one and a half developers freed and about $30,000 a month in savings. The page does not say how the savings were measured.
Full details are in the original sources.
