JD Health Unveils Jingyi Qianxun 3.0, Flags Three Gaps in AI Healthcare

The 2026 JD Global Technology Explorer Conference was held in Beijing on September 9. At the AI + Health themed forum, Xu Jun, vice president of JD Health and head of its technology and product department, presented the company's in-house medical large model, Jingyi Qianxun 3.0.

Xu said AI healthcare can already handle functions such as consultations and report interpretation, but three business gaps remain. The first is a demand recognition gap: health data is non-continuous and AI lacks long-term memory, so it tends to rush to a judgment when information is incomplete. The second is a service continuity gap: after an abnormality is identified, there is a lack of follow-up testing, diagnosis and treatment, and medication services, so only half the problem is solved. The third is a long-term management gap: health management is a long-cycle process, while most existing AI healthcare offerings remain limited to single question-and-answer exchanges and lack continuous tracking and management capabilities.

Xu proposed that cross-scenario and long-duration capability are the two core keywords for the next stage of AI healthcare development. Cross-scenario means connecting online and offline, in-hospital and out-of-hospital settings, and achieving capability integration across multiple terminals such as apps and smart wearable devices. Long-duration means changing the model in which AI healthcare is confined to single question-and-answer exchanges, and instead providing users with continuous, full-cycle health management services along time dimensions of days, months, and even years, supported by long-term memory capabilities.

Xu said that to achieve these two leaps, Harness is particularly important. Harness originally means horse tack, including reins, saddle, bit, and horseshoe. In 2026, the concept began to gain popularity in the artificial intelligence field, referring to the entire supporting system built for an AI model, which determines which tools the model can call, which resources it can access, how information flows between different sub-agents, and when execution terminates. Xu defined Harness as the action space for AI healthcare, and said this action space will be the engineering vehicle for bridging the three business gaps. He said that how to schedule various professional resources within this action space and integrate professional medical services into the joint training system of AI is a mandatory question for the industry.