Digital China Tang Kai: The breakthrough point for deep integration of AI + healthcare lies in high-value segments

【Global Times—Technology News Report, reporter Lin Mengxue】From 2022, when large-model technology suddenly emerged, to 2025, when various kinds of agents have congregated in large numbers, AI+Healthcare has also evolved from its early days of medical imaging recognition to medical agents being deeply embedded in the core clinical workflow—moving toward the critical step from “works” to “works well.”

During the 2026 Zhongguancun Forum annual conference, Tang Kai, Vice President and Chief Engineer of Sinocan Digital, said that in the integration process of AI+Healthcare, the real test of this path is whether “AI can deliver a response that is truly trustworthy in the places where doctors need it most.”

“The essence of the last mile”_ is a value question_

“The last mile of AI+Healthcare is, in essence, a value question. The key is whether AI can create real value for doctors. If it can’t provide value, everything will easily turn empty.” Tang Kai said. “The moment when AI breaks through the loop will inevitably land in high-value areas. In clinical settings, high value is especially reflected in clinical challenges such as ‘acute, critical, severe, hard-to-diagnose, difficult, and rare.’”

Tang Kai believes that this year, the application path of AI in the medical field has become increasingly clear, especially as the implementation pace of technologies represented by large models has significantly accelerated. The core marker of this shift is that the R&D and application of medical agents have become a strong wave of industry development, and the focus of development is moving from relatively independent agents toward more complex, deeper business scenarios. “Since 2025, AI has begun to create important value at multiple key nodes in medical workflows. Sinocan Digital, along with institutions such as Peking Union Medical College Hospital, is co-creating in depth. We plan to jointly advance intelligent diagnosis and treatment systems such as MDT (multidisciplinary diagnosis and treatment) based on large models, to assist clinical decision-making for difficult and complex cases. This also shows that AI technology is continuously penetrating the core business areas of healthcare.” he said.

In pursuing AI+Healthcare, Sinocan Digital upholds the core philosophy of “AI for Process,” meaning that artificial intelligence is deeply integrated into business processes to create tangible value. This philosophy has also become a key guiding principle for its technology deployment: “In the medical field, we strictly follow the advancement of medical business flows. Currently, our focus centers on the hospital’s core diagnosis and treatment processes, covering the entire perioperative workflow from pre-op, intra-op, to post-op. Based on this, we develop a series of intelligent applications.” Tang Kai explained. At present, solutions such as diagnosis and treatment of postoperative complications and preoperative anesthesia assessment have been deployed and used in hospitals, and the deployment of this series of applications is a vivid practice of the “AI for Process” philosophy.

Tang Kai said, “Sinocan Digital has cooperated with Peking Union Medical College Hospital to develop an agent for postoperative complications of pancreatic cancer. It can quickly identify the risk of complications, save doctors nearly 80% of their time, and the accuracy has stabilized at above 94%.”

And for doctors to be willing to use this intelligent agent proactively, the core stems from two major real values: “First, it can help doctors cross-validate diagnosis and treatment judgment results, thereby reducing the rate of misdiagnosis; second, it can significantly improve work efficiency.” Tang Kai further pointed out that the deployment of AI in healthcare is not constrained by technical challenges; more importantly, it must be able to achieve “a small interface with a big result,” meaning that by making a lightweight technical entry point, it can produce remarkable clinical effectiveness. Based on this, he summarized three progressively deeper layers of value that AI creates for healthcare: “The first layer is performance value—through intelligent agents, AI improves doctors’ efficiency and work quality. The second layer is decision value, which is an important direction for the future. Doctors’ day-to-day core is decision-making, and whether intelligent agents can become a reliable decision-support role is a major challenge. The third layer is discovery value—through deep collaboration with hospitals, exploring more cutting-edge areas such as the diagnosis and treatment of difficult diseases. Only by creating value truly within the diagnosis and treatment workflow can we more thoroughly open up the ‘last mile.’”

“Data” is the mountain that must be crossed

“As applications go deeper, we find that the key difficulties are not AI or large-model technology, but data.” Tang Kai said plainly. “The quality of data and the completeness of processing workflows will directly determine the depth and sustainability of AI applications.”

To this end, this year Sinocan Digital has started to actively cooperate with hospitals to jointly build high-quality medical datasets. “We are exploring the construction of high-quality disease-specific datasets, centered on various diseases.”

In Tang Kai’s plan, Sinocan Digital will adhere to a “two-track” approach to laying out AI+Healthcare: “First, we will continuously deepen at the application layer, pushing AI to play a greater role in core diagnosis and treatment areas. Second, we will strengthen the foundation at the data layer, supporting ‘AI for Process’ with ‘Data for Process.’ This is a path that requires long-term investment, and we will keep moving forward in this direction.” And this layout is also deeply aligned with today’s popular technology concepts of “service twins” and “multi-agent coordination.”

Regarding the technical development stage of “service twins,” Tang Kai holds a generally positive and optimistic view, believing that it has now entered the stage of engineering practice.

But he also pointed out that the development of service twins must cross the “data” mountain. “The development of service twins—or, the advancement of ‘AI for Process’—in essence is a data problem. Currently, across many industries, data quality still often falls short of supporting deep implementation of applications like this.”

Tang Kai used “digital engineering” in manufacturing as an analogy: “In manufacturing, we are pushing for ‘digital engineering.’ The core is to build digital twins at the data layer and form a precise digital portrait model of equipment. Only by achieving seamless connection at this layer can we carry out more designs on the service-twin layer.” In the healthcare field, Sinocan Digital’s exploration direction is very clear: “Focusing on disease-specific conditions, we build a fundamental engineering model for the diagnosis and treatment process, and we drive ‘AI for Process’ toward maturity through ‘Data for Process.’” He added, “We hope that in the future we can achieve highly coordinated working scenarios among ‘digital humans, robots, and bio-humans (doctors).’”

“From a doctor’s perspective, in 2026 they will gradually start to feel that AI truly enters the working process, and there will be more and more such agents. However, standing on the patient side, the AI-enabled experience that can be clearly felt during the process of seeking care is still limited at the moment, and this will also be a key direction for deepening applications in the next stage.” Tang Kai said.

A massive amount of information, precise interpretation—everything is available in the Sina Finance APP