Unisound has recently entered into a partnership with a leading property and casualty (P&C) insurance group on a fund oversight system project. Drawing on Unisound’s proprietary U2-Med large language model for healthcare, medical insurance and pharmaceuticals, intelligent audit algorithms, and medical insurance expertise, the two parties will jointly develop an integrated oversight solution covering early warnings, monitoring during operations, and retrospective reviews. The solution will help identify irregularities and potential risks in fund use, supporting the secure, compliant and efficient operation of medical insurance funds.

Unisound previously won contracts for two key projects with the same insurance group: a Medical Insurance Fund Oversight Agent and a Medical Insurance Knowledge Q&A Agent. These projects have enabled the company to continue exploring applications of AI agents in medical insurance oversight and public services. The latest agreement marks another application of Unisound’s specialized medical AI capabilities in strengthening the healthcare security system. It will support more intelligent and precise oversight of medical insurance funds and extend the application value of large language models in this field.

Building Integrated Intelligent Risk Controls Across the Fund Use Lifecycle

Oversight of medical insurance funds spans a long and complex process involving policy rules, clinical care and expense settlement. As medical insurance policies evolve, clinical practices vary, and business data volumes grow, oversight teams face persistent challenges in efficiently identifying risks, pinpointing anomalies, and providing reliable evidence for on-site investigations.

Traditional review processes rely largely on fixed rule engines and manual checks. In complex cases that require cross-analysis of medical records, diagnoses, prescriptions, tests and itemized expenses, matching individual indicators rarely captures the full clinical reasoning behind care decisions. This can lead to missed investigative leads and add to reviewers’ workloads.

To address these challenges, Unisound combines its specialized medical large language model with intelligent audit algorithms and connects medical knowledge bases, medical insurance policy databases, and business rule repositories. The solution performs relational reasoning across heterogeneous data sources, including clinical documentation and settlement records, to uncover potential risks and provide reviewers with high-quality investigative leads and supporting evidence.

The solution covers the entire fund use process. It issues risk warnings at the outset, detects anomalies as operations proceed, and consolidates evidence afterward to support retrospective investigations. This approach helps medical insurance oversight move toward proactive, intelligent controls across the full lifecycle, reducing reliance on reactive reviews after the event.

Strengthening U2-Med to Connect Clinical Reasoning with Medical Insurance Rules

Medical insurance oversight requires both an assessment of whether expenses comply with payment rules and an understanding of the clinical reasoning behind care decisions. Determining whether a test, prescription or charge is appropriate often requires a combined analysis of the patient’s condition, the evidence supporting the diagnosis, and relevant policies.

Built on Unisound’s proprietary U2-Flash foundation model, U2-Med incorporates specialized knowledge across healthcare, medical insurance, including commercial insurance, and pharmaceuticals. It can understand professional content such as medical records, examination reports, prescriptions and policy provisions, and perform information extraction, relational analysis and assisted reasoning. These capabilities support medical insurance reviews, commercial insurance claims processing and related tasks. In OpenAI’s HealthBench Full and HealthBench Hard evaluations of medical dialogue capabilities, U2-Med scored 73.67 and 61.07, respectively, ranking first among the models included in this evaluation on both benchmarks.

To help medical AI progress from isolated question answering to understanding care as a continuous process, Unisound has also developed U2-Med-World, a medical world model. By reconstructing real-world care trajectories and generating simulated clinical scenarios, it enables U2-Med to learn through interaction how to ask proactive follow-up questions, update its assessments as new information emerges, and identify potential risks.

In fund oversight, U2-Med’s specialized understanding helps connect medical knowledge with medical insurance rules. It analyzes how expenses relate to payment rules while examining the relationships among diagnoses, treatments, tests and medications. This supports the identification of potential anomalies and improves the quality of investigative leads.

Across medical knowledge understanding and medical insurance analysis, Unisound continues to integrate large language model capabilities with industry rules and business processes so that specialized AI can better support practical operatio

Medical insurance operations involve diverse data types and tasks. They require an understanding of medical records, policies and expenses, as well as specialized analysis of supporting materials such as medical images. Unisound continues to develop its medical AI technology and practical experience to meet these needs.

In medical imaging, the Guiyi Guidian Unisound Joint Innovation Team and the X Vision Explorers Team, which Unisound helped establish, won first prizes in the Proprietary Model and Self-Built Model categories, respectively, in the “Intelligent Detection of Multiple Diseases Based on Chest X-rays” track of the 2026 National Medical Insurance Imaging AI Recognition Competition. The awards reflect the company’s technical expertise in medical image recognition and multitask coordination.

In medical insurance settlement, Unisound and Peking Union Medical College Hospital jointly developed a large language model-based solution for intelligent generation of diagnosis-related group (DRG) settlement lists. The solution won first prize in the finals of the 2024 National Smart Medical Insurance Competition. Covering medical record understanding, coding and settlement list generation, it explores the application of large language models in medical insurance payment management.

Through its work in medical image recognition and medical insurance settlement management, Unisound continues to align AI technology with practical operational needs, building technical expertise and implementation experience for intelligent healthcare security systems.

Expanding AI Applications Across the Three Domains to Advance Healthcare Security

Clinical services, medical insurance payments and insurance protection each serve distinct purposes, while sharing requirements for medical knowledge understanding, risk identification and operational capabilities. Applying specialized AI to these different settings can help address needs across healthcare and coverage services.

The partnership with the leading P&C insurance group extends Unisound’s work in applying its large language model for healthcare, medical insurance and pharmaceuticals, together with its industry experience, to medical insurance fund oversight. Looking ahead, Unisound will continue to advance applications of large language models, multimodal AI and AI agents, integrating specialized AI capabilities into relevant business processes to support the secure use of medical insurance funds and improve the quality and efficiency of healthcare security services.