Generative AI (GenAI) is already revolutionizing healthcare systems both in Israel and globally, from enhancing diagnostic procedures to managing patient experiences and overseeing complex operational processes. Leumit Health Services, in collaboration with Matrix DnA, is among the first organizations in Israel to implement GenAI in the oversight of medical documents coming from hospitals—a process that until recently was carried out manually by doctors and medical teams. The infrastructure established will enable a wide range of additional applications, dramatically improving service for the health fund’s members.
The Challenge of Manual Medical Document Oversight
Today, when a patient undergoes a procedure at a hospital, they receive documents such as Form 17, which is pre-approved or retrospectively approved by the health fund. Once the patient submits the form to the hospital, the payment matter is considered closed from their perspective. However, for the health fund, which is responsible for ensuring the well-being of its members and the efficient use of public funds, the work is just beginning. Without an organized interface with hospitals, oversight teams—including doctors and specialists—must review each set of documents, compare the information submitted by the hospital with the patient’s data, and ensure full alignment between charges and the procedures actually performed. This manual process consumes significant resources: valuable doctors’ time, high staff burnout, and an inherent risk of human error that could lead to substantial financial losses.
The GenAI Solution – Automated and Accurate Oversight
Leumit Health Services chose to implement an innovative GenAI-based solution developed by Matrix DnA, which allows automatic and precise extraction and comparison of information from medical documents. The system can identify valid charges that do not require manual review, saving time and resources. It detects irregular or incorrect charges and provides clear, reasoned explanations, enabling quick error correction and accelerating the appeals process when necessary. Additionally, the system identifies potential appeals even in documents that have not been manually reviewed, uncovering funds that might otherwise be overlooked.
How It Came About
Harel Shrabi, Chief Financial Officer, envisions increasing efficiency in the public system to free resources for improving medical quality and expanding service capabilities based on resources saved from correcting erroneous charges. With a clear vision, Harel recognized the potential of artificial intelligence in achieving these goals.
Connecting Legacy Systems with Technological Innovation
One of the main challenges of the project was integrating outdated legacy systems with innovative technology. For example, extracting information from old legacy systems and processing it to be understandable and precise for the LLM model. Special attention was given to reliability, ensuring that the system not only identifies discrepancies but also explains findings in professional medical language. To achieve this, the solution incorporated multiple models—one focused on comparison and oversight, and another on providing clear, professional explanations familiar to doctors.
The automated interface built between hospitals and Leumit Health Services creates a technological infrastructure that can serve as the foundation for a range of new services. In the future, this could significantly improve patient experience—for example, through full transparency in procedures, charges, and medical processes.
Rakefet Yaakobi, CIO of Leumit Health Services: “We are proud to be at the technological forefront of Israel’s healthcare system. GenAI technology allows us to do much more to optimize and improve healthcare operations. Implementing an automated medical document oversight system, developed in close collaboration with Matrix DnA, enables us to allocate medical staff to higher-value tasks and significantly enhance operational efficiency. As a result of this project, we expect to save millions of shekels annually. These savings allow us to allocate additional resources for the benefit of our members and other areas.”
The organizational decision to start a project that does not directly affect frontline patients proved effective. Harel advocated beginning on the administrative side to minimize risks and allow rapid progress. Meanwhile, Leumit is already developing other GenAI-based applications, using AI extensively to efficiently advance other innovative projects.
Matrix DnA: “We are excited to participate in a significant project that leads to process efficiency, resource savings, and reduced pressure on medical teams. We worked closely with Leumit’s team to deeply understand the complex oversight needs, and together built a flexible and precise solution capable of extracting data from complex medical documents, performing comparisons, and generating explanations in professional medical language. Our collaboration with Leumit Health Services demonstrated how GenAI technology can solve real problems and create meaningful value in the healthcare system.”