JINXIN Fertility Launches "IVF Co-pilot," China's First Clinical AI Agent for Assisted Reproduction

At a fertility clinic in Chengdu, a new kind of colleague has started its first shift. It doesn't wear scrubs or carry a stethoscope. It lives inside the hospital's electronic medical records system, and its job is to help doctors decide which ovarian stimulation protocol to use for in-vitro fertilization patients.

The system, called IVF Co-pilot, was developed in-house by Jinxin Tech — the technology arm of JINXIN FERTILITY (HKEX: 01951) — and has begun piloting at Sichuan Jinxin Xinan Women and Children's Hospital's Bisheng campus. It is, according to the company, the first clinical AI agent designed specifically for fertility doctors in China.

The launch marks a notable step in the application of artificial intelligence to one of medicine's most data-intensive and deeply personal fields. Assisted reproduction demands that physicians weigh ovarian reserve markers, baseline hormone levels, prior cycle responses, and real-time ultrasound and lab results — often in a matter of minutes — to tailor a treatment plan for each patient.

"The clinic doesn't lack information," said Wang Weipeng, head of research and development at Jinxin Tech. "The real challenge is how to extract actionable insights from a vast history of past cycles and professional literature within the limited time available, so the attending physician can make better-informed decisions."

Data Advantage

Jinxin Fertility's scale gives the system a head start. According to Frost & Sullivan's 2026 China Assisted Reproduction Industry White Paper, Jinxin Fertility is the country's largest integrated assisted reproduction services platform by oocyte retrieval cycle volume. In 2025, the company performed more than 23,000 cycles across China, bringing its cumulative total past 300,000 oocyte retrieval cycles. Fewer than 20 fertility clinics nationwide exceed 5,000 cycles per year.

A separate Frost & Sullivan report, the Global Assisted Reproduction Report 2026, ranked the success rates of China's leading platforms among the highest in the world.

"The massive volume of cycle data and successful cases is the critical foundation for developing this system," Wang said. "In compliance with applicable laws and data security requirements, the system can screen treatment protocols and relevant data from similar cases in professional databases, then use a locally deployed large language model to summarize, compare and assist analysis — providing a reference for clinical decision-making."

How It Works

The IVF Co-pilot currently focuses on assisting with ovarian stimulation protocols and initial medication dosing. It integrates hospital clinical experience, IVF domain knowledge, clinical algorithms and a locally hosted large language model directly into the electronic medical records workflow. The output is designed to be comparable, traceable and reviewable by physicians.

"On one hand, the system draws on the IVF knowledge base and the hospital's historical cycle data to provide doctors with professional knowledge and clinical experience references," Wang explained. "On the other hand, clinical algorithms and the local large language model jointly analyze both types of information to produce comparable, traceable opinions that doctors can review — presented clearly within their existing workflow."

Wang likened the system to a co-pilot in a commercial aircraft cockpit or a rally co-driver reading pace notes: "It doesn't replace the driver, but it continuously provides the support needed. For clinical experts, it surfaces key metrics and supporting evidence at the right moment. For junior doctors, it offers hospital-reviewed, traceable best-practice references. All treatment decisions remain with the physician, based on the complete medical record and the patient's individual circumstances."

Trust, Not Mimicry

The approach reflects a broader shift in how medical AI is being built — away from models that try to sound like doctors, and toward systems that fit into how doctors actually work.

"What we are building is not a model that 'sounds like a doctor,'" Wang said. "It is an AI capability that can be used, examined and trusted by doctors. Whether medical AI creates value depends not only on the model's capability, but also on whether it appears at the right point in the workflow."

If the pilot succeeds, the question will not be whether AI can practice medicine. It will be whether enough clinicians are willing to let a machine sit in the co-pilot's seat.