GC MediAI’s Uisarang AI used by about 800 medical institutions

At a demonstration in Seoul’s Yeouido district, a diabetes consultation began and the conversation between the doctor and patient was immediately converted into text. A medical record appeared on the screen, organizing glycated hemoglobin and fasting blood glucose levels, dizziness and changes to medication times. This was GC MediAI’s medical artificial intelligence service, Uisarang AI. About one month after its launch, it is being used by about 800 medical institutions, including existing and external customers.

Uisarang AI records conversations between medical staff and patients in real time and organizes them in the SOAP format. SOAP is a medical-recording method that divides entries into subjective symptoms, objective test results, assessment and treatment plan. The service summarizes previous medical records, medications, testing intervals and family history, and supports searches for diagnosis and prescription codes corresponding to the conversation. Once a consultation ends, it identifies patients who need follow-up visits and recommends the wording and timing of patient-specific notification messages.

Kim Seok-jung, GC MediAI’s chief product officer, explained that using an artificial intelligence service not connected to an electronic medical record requires at least eight operations per patient, including entering patient information, checking previous records, switching screens and copying results. Uisarang AI is designed to automate this process while having medical staff make the final check before incorporating the results into the electronic medical record. It also includes a function for asking in natural language about operating indicators such as hospital revenue and patient numbers and checking them in real time.

Accuracy and security are among the company’s key selling points. The medical-specialized speech recognition model developed in-house recorded 97.1% accuracy in recognizing medical terminology in the company’s own comparative evaluation. OpenAI’s Whisper large-v3 and Alibaba’s Qwen3-ASR were each measured at 90.5% for comparison. The company said the character error rate was 5.56% and the word-segment error rate was 15.80%. These figures are the results of GC MediAI’s own evaluation.

So what changes in practice? Medical staff can reduce the need to enter charts separately during consultations or move records between multiple screens, allowing them to focus more on conversations with patients and reviewing results. Hospitals can handle follow-up guidance and check operating indicators within a single screen. GC MediAI plans to collect usage data and feedback from the field by offering the service free of charge through the end of this year, then launch the cloud-based electronic medical record system Uisarang Cloud in March next year. It plans to subsequently expand the service into a medical operating-system platform connecting pharmaceutical, healthcare, medical device, telemedicine, financial and online purchasing services.