China News Service, Changchun, September 7 (Reporter Guo Jia) To complete the data management of 35,000 business tables, the traditional method is expected to require a month's workload of about 1,152 people; now with the help of intelligent agents, this workload has been compressed to a month's workload of about 12 people, manpower input is reduced by more than 95%, and the overall processing efficiency is increased by more than 30 times.
This set of data comes from the First Hospital of Jilin University and is a microcosm of artificial intelligence moving from “auxiliary tools” to “digital employees”.
"The real value of AI is not just what the model can answer, but whether it can be incorporated into actual business processes." Zhao Zhenwei, director of the Information Center of Jilin University First Hospital, said in a recent interview with a reporter from China News Service.
At the end of July this year, the Medsuper Harness multi-agent collaborative work platform independently developed by the hospital's information center team was deployed and launched on the hospital's intranet. Different from traditional chat-based AI, this platform allows AI to undertake complete business work: understanding needs, calling tools, and executing tasks, which are promoted collaboratively by agents with different roles.
In data governance, intelligent agents can independently complete demand understanding, table field identification, data extraction, cleaning conversion and quality verification. In addition to increasing the efficiency of processing repetitive data requirements by more than 90%, the overall business processing efficiency has increased by more than 30 times.
AI doesn’t just enter the hospital’s data backend. In the field of software research and development, the platform sets up agents with different roles such as product managers, architecture design, development, testing, and operation and maintenance, covering the entire process of demand analysis, prototype design, code development, testing and verification, and deployment and delivery. Data shows that the development cycle of conventional applications is shortened by more than 50% on average, the efficiency of demand analysis and prototyping is increased by more than 3 times, and the efficiency of repetitive coding, interface development and testing is increased by more than 80%. The development cycle of some lightweight applications is shortened from weeks to days.
For hospitals, this means that some clinical and management needs that required repeated communication and queued development in the past can be transformed into practical applications faster.
Zhao Zhenwei believes that a hospital is a highly complex digital organization. Doctors, nurses, pharmacists, engineers and managers use different information systems, and a large amount of work relies on data and system collaboration. It is difficult for a single chatbot to undertake these complex tasks, but multi-agent agents can break down a job into multiple links, and then complete it collaboratively by different AI roles.
Clinical scenarios are also being explored. The First Hospital of Jilin University has tried to use intelligent agents for patient grouping, anesthesia assessment and intraoperative reminders. The intelligent agent can combine the patient's medical history, examination and monitoring data to prompt risks according to set rules, but the final judgment is still made by medical staff.
What has received equal attention as efficiency is the safety boundary of medical AI. The platform is deployed on the hospital intranet, with relevant models, knowledge bases and business interfaces running within the hospital, and manages data through identity authentication, authority control, data desensitization and operational auditing.
In Zhao Zhenwei’s view, the relationship between hospitals and AI is changing from informatization, digitalization to intelligence. In the past, people needed to learn how to operate the system; but as artificial intelligence moves from "answering questions" to "being able to get things done," the digital construction of hospitals is entering a new stage.
