Specialized AI Agents Drive a Critical Leap in Health Management: From Reactive to Proactive, End-to-End Innovation

In recent years, AI-driven health management has been undergoing a profound transformation from “reactive response” to “proactive prevention.”

From the emergence of OpenAI in late 2022, which revealed the potential for deploying large language models in health management applications, to the implementation of AI Agents in 2025 that infused health management with greater intelligence, and further to the viral popularity of “Lobster” in early 2026 that ignited interest in exploring local agent deployment, each technological leap has provided solid support and boundless possibilities for the advanced development of health management.

In this process, AI-driven health management has long ceased to be a passive responder engaged in simple “question-and-answer” exchanges, nor is it merely a single-point tool that “disappears” after service delivery. Instead, it has evolved into a collaborative partner capable of proactive interaction and providing comprehensive management across the entire care continuum—pre-diagnosis, during diagnosis, and post-diagnosis. Furthermore, the impact of AI on health management extends beyond mere functional iterations; it represents a paradigm shift from “passive response” to “proactive prevention.”

However, to be frank, this paradigm shift remains incomplete. Core challenges such as data silos, model interpretability, and multi-party collaboration among hospitals, enterprises, and insurers still await resolution. In other words, there is still a long way to go for AI-driven health management to evolve from merely “functional” to truly “effective.”

To this end,The 19th Transaction Roundtable of the China Innovative Medical Assets Living Room, co-hosted by VCBeat and MicroMed, featured two distinguished guests: Li Jiaming, Deputy General Manager of Lejian Health Technology Group, and Wu Lei, AI R&D Director at E-Health Now. Centered on the theme “AI Reconstructing Health Management: Full-Chain Innovation from Passive Delayed Intervention to Proactive Instant Prevention,” they provided an in-depth deconstruction of the current landscape, key application scenarios, and future development trends of this paradigm-shifting revolution in AI-driven health management.

Throughout the integrated development of AI and health management, different periods have featured distinct focal points. Addressing this, Li Jiaming, Deputy General Manager of Lejian Health Technology Group, and Wu Lei, AI R&D Director at E-Health Now, shared their respective observations from the perspectives of application and technology during a live broadcast.

Li Jiaming first outlined, from an application perspective, three fundamental changes that AI has brought to health management: shifting from “reactive response” to “proactive prevention,” from “single-point tools” to “full-chain integration,” and from “standardized outputs” to “personalized interventions.” In his view, the core value of these changes lies in “enhancing efficiency.”—This enables healthcare providers to serve an exponentially growing user base with the same workforce, thereby freeing up medical staff’s time and energy to focus on more core and challenging tasks. Furthermore, by leveraging the capabilities of large language models, enterprises can enhance their personalized and precise health management services, which in turn improves user adherence and service experience.

Wu Lei outlined the developmental trajectory of AI-driven health management from the perspective of technological evolution.: The emergence of GPT models in 2022 revealed the potential of large language models (LLMs) to empower the development of health management services, sparking a wave of integrated development combining LLMs with health management. At that time, explorations primarily focused on data analysis-based applications, such as medical report interpretation. This phase was characterized by isolated, fragmented, and “passive response” solutions. Around 2024, the industry officially entered a stage of large-scale LLM deployment, with numerous point-specific applications of “LLM + Health Management” flourishing, thereby laying the groundwork for the subsequent “integration” of end-to-end processes. Since then, the industry has been exploring intelligent applications across the entire health management workflow on one hand, while on the other, leveraging the rise of AI agents to transform health management services from “passive response” to “proactive early warning,” ultimately achieving truly proactive, full-cycle, and end-to-end health management.

Based on the insights shared by the two guests,AI is transforming health management from a passive service model of “user-initiated, system-responsive” into a continuous companionship characterized by “system monitoring and proactive intervention.” This represents not only an enhancement in technological capabilities but also a fundamental restructuring of service logic.

Beyond these three structural shifts—from “reactive response” to “proactive prevention,” from “single-point tools” to “end-to-end penetration,” and from “standardized output” to “personalized intervention”—the deepened application of large language models also presents health management companies with a strategic choice.

Amid the significant enhancement of AI capabilities, health management companies face a fundamental strategic choice: Should they build a comprehensive generalist platform covering a wide range of services to address “all health issues” for “everyone,” or should they focus on specific diseases to provide in-depth specialized disease management services?

Li Jiaming first clarified Lejian’s judgment: the transition from general practice to specialization is an inevitable path for the development of AI-driven health management.Therefore, Lejian has taken the lead in strategic deployment by launching an AI-powered chronic disease management system for the “Five Highs” (hypertension, hyperlipidemia, hyperglycemia, hyperuricemia, and overweight/obesity). This system not only provides personalized dietary plans but also offers 24/7 online analysis of abnormal data, enabling real-time response and intervention for irregularities.

Lejian’s strategic focus on managing “five highs” through specialized disease management is driven by three key considerations: First, disease correlation—approximately 70% of chronic diseases are associated with overweight or obesity. Weight loss serves as the “infrastructure” for chronic disease management, offering a leverage effect where addressing one issue positively impacts the entire system. Second, data foundation—built upon long-term accumulation of tens of millions of anonymized health records, Lejian’s self-developed Tianrui Qiyuan Medical Large Language Model has achieved a 95% risk warning rate in the “five highs” domain, providing a robust data backbone for deep specialization in these conditions. Third, service closed-loop—Lejian has established an integrated pathway combining “health check-ups + chronic disease management + home nursing services,” enabling specialized disease management to span the entire process from detection and intervention to rehabilitation, thereby avoiding the fragmentation characterized by “diagnosis without intervention” or “intervention without follow-up.”

“From generalized health management to precise disease-specific management is not only an upgrade in technical capabilities, but also an evolution of the business model.”“Only by delving deep and thorough can enterprises forge genuine differentiated advantages,” said Li Jiaming.

Wu Lei, while acknowledging this trend, deconstructed its underlying logic from four dimensions.First, the logic underlying the demonstration of clinical value dictates the inevitability of specialization. While general practice addresses the issue of coverage breadth, it is deep intervention at the level of specific diseases that truly improves patient health outcomes and reduces healthcare expenditures.Secondly, the demands of payers are escalating—insurers, banks, and large enterprises are more willing to pay for quantifiable outcomes rather than generic health advice. Furthermore, the natural evolution of data collection will “automatically” drive specialization.As service volume accumulates, multidimensional data naturally converges along specialty-specific dimensions, creating a positive feedback loop.Ultimately, user trust is built on perceived outcomes.—Lower blood sugar and reduced weight are far more compelling than generic advice.

However, the transition from general practice to specialized care is far from smooth. Wu Lei bluntly stated that the biggest bottleneck during this process lies in data.Disease-specific management requires access to more in-depth medical data—diagnostic records, medication regimens, test results, and disease progression trajectories. However, such data are highly concentrated within hospitals and health authorities, and the restriction that internal hospital data must not leave the institution represents a widespread hard constraint across the industry. In response, Wu Lei pointed out that exploring compliant ways to collaborate with hospitals and achieve joint modeling while safeguarding privacy is a challenge the industry must collectively address.

Li Jiaming added another challenge: the interpretability of the model.In general practice scenarios, users have a higher tolerance for AI errors; however, in specialized disease scenarios, every AI judgment may be critical to patient safety. This requires AI not only to provide conclusions but also to offer the underlying rationale—citing which clinical guidelines were referenced, which indicators were used as the basis, and which differential diagnoses were excluded. Li Jiaming mentioned,In its practice within the “Five Highs” domain, Lejian has consistently prioritized “interpretability” as a core metric for model iteration, ensuring that both users and physicians can understand the AI’s reasoning process.

To this day, the question of “who pays for health management” continues to constrain industry development. Within China’s healthcare system, basic medical insurance primarily covers disease treatment; a significant proportion of individual consumers lack the habit and awareness of paying for health management services, resulting in low willingness to pay out-of-pocket; corporate procurement by B-side enterprises is constrained by budget cycles and employee benefit policies, making it difficult to establish stable, scalable revenue streams... In response,Commercial health insurance, as an innovative payment model, is regarded as the key direction for breaking the deadlock.

Li Jiaming analyzed the advantages of positioning health insurance as a payer from three perspectives:For individual consumers, improving their physical condition through health management not only yields health benefits but also allows them to enjoy premium discounts or rebates, creating a positive incentive loop. For insurance institutions, health management facilitates a shift from “probability-based gaming” to “value management.” The profitability model of traditional insurance is built on the “law of large numbers”—earning a spread between premiums and claims by actuarially predicting the probability of payouts. With the introduction of health management, insurers can proactively intervene in users’ health status, reduce disease incidence and medical expenditures, thereby lowering claim ratios, enhancing profitability, and improving customer stickiness. For health management companies, users exhibit higher compliance and willingness to utilize services due to premium incentives, with health insurance acting as a payer that supports the sustainable operations of these enterprises. Therefore, “the integration of health insurance and health management is an inevitable trend, resulting in a win-win-win situation for users, insurance institutions, and health management organizations,” summarized Li Jiaming.

Yet the path to integration is far from smooth. Wu Lei pointed out two major practical challenges.First, the penetration rate of commercial health insurance remains low, and coverage needs to be strengthened. Many consumers’ understanding of commercial health insurance is still limited to traditional products such as “critical illness insurance” and “million-yuan medical insurance,” with little awareness of new models that integrate “insurance + health management.” Second, data barriers persist. Insurance companies maintain strict underwriting and policy issuance protocols, while hospitals must comply with medical data regulations. As intermediaries, health management companies have limited flexibility to make dynamic adjustments, which may result in suboptimal health management outcomes.

In the face of these challenges, the industry is exploring multiple pathways to break through. First, in terms of data interoperability, joint modeling is being achieved through technologies such as federated learning and privacy-preserving computation, enabling “data stays put while models move” to facilitate value co-creation without disclosing raw data. Second, regarding system co-construction, new collaborative models such as “payment for performance” and “risk sharing” are being explored to deepen partnerships among insurance companies, hospitals, and health management enterprises. Third, in payment innovation, efforts are underway to include health management services in the reimbursement lists of basic medical insurance or commercial health insurance, striving for institutional breakthroughs at the policy level. Although these explorations are still in their early stages, the direction is clear.

As AI-driven health management continues to evolve, Wu Lei and Li Jiaming have each offered their perspectives on future trends in the field, including product forms, technological boundaries, and social value.

In terms of product form, Wu Lei believes that specialized AI agents will experience a concentrated boom in the next 1-2 years, with a large number of vertical products rapidly emerging on the market. These agents will cover multiple aspects including screening, early warning, intervention, and follow-up.

Meanwhile, Li Jiaming stated,In the past, by reducing information asymmetry in health knowledge and disparities in healthcare resources, AI has enabled more people to equitably access basic health services. This signals that AI has successfully validated the first half of its practical value in implementing health management. AndIn the second half, a tiered service system will naturally emerge.—The free basic tier meets the public’s everyday consultation needs, the paid professional tier offers in-depth intervention and management, and the premium VIP tier combines human experts with dedicated services. This tiered structure is not a simple stratification where “the wealthy enjoy better service”; rather, it leverages AI to reduce marginal costs, enabling basic services to reach a broader population while delivering higher-value services to users with more complex needs.

Finally,The two guests highlighted several key development directions worthy of attention in the future: first, the integration of rehabilitation and elderly care with embodied intelligence.With the development of embodied AI robots, the long-marginalized field of rehabilitation and elderly care is set to welcome new technological variables.The second is the digital reconstruction of Traditional Chinese Medicine.The empiricism and non-standardized nature of Traditional Chinese Medicine (TCM), accumulated over thousands of years, have become ripe for re-encoding in the era of large language models (LLMs). Classical texts, medical case records, and herbal formulas constitute a rich repository for LLMs to learn from.Third, AI's exploration from virtual environments to the physical world.Future AI-driven health management will not be limited to mobile apps; it will also integrate with automated laboratory equipment, wearable devices, and even surgical robots, enabling an end-to-end automated workflow from “pre-consultation” to “pre-testing” and then to “pre-intervention,” thereby significantly enhancing service efficiency and user experience.

Overall, from single-point tools to full-process intelligent assistants, from passive response to proactive early warning, and from general practice to specialized care—In 2026, AI-driven health management stood at a critical juncture, transitioning from “functional” to “user-friendly.”The possibilities of technology have been unlocked, while the adaptation of the industrial ecosystem and the integration of data... will determine the ultimate depth of this transformation.