When Every Doctor Has an AI Superteam, How Will Diagnosis and Treatment Change?

A notice posted at a hospital registration desk has recently sparked heated discussion on social media.

The notice reads: "If you have already obtained a diagnosis from an AI model, please do not seek our opinion again; please consult another AI model instead."

Behind the joke lies the real situation facing medical AI today. Patients walk into the consulting room holding an AI diagnosis; doctors look at the AI's conclusion and fall silent. What exactly is the relationship between AI and doctors? How should "AI + hospital" actually be done?

There is no unified answer to this question within the healthcare industry itself. Medical practice and the settings in which it occurs are in themselves a highly interconnected, complex system. Single-point AI tools are proliferating — image recognition, diagnostic assistance, consultation chatbots — with products at almost every step. Yet products that truly enter hospitals' core clinical workflows and are used daily by doctors across the whole process are few and far between.

The difficulty that single-point tools have in entering core workflows is sometimes a dispute not about the technology itself, but about liability, habit, trust and other factors.

Against this backdrop, the solution for "AI + hospital" may well be to embed AI into the clinical workflow so that it becomes a decision-support collaborator. New approaches are already emerging in the industry. For example, the Agent Hospital system developed by Tairex seeks to break out of the limitations of single-point tools and build a multi-agent collaborative medical ecosystem.

Tairex's answer is that once human doctors "teach by example" their avatar agents, those agents become the doctors' assistants.

When a patient with hyperthyroidism walks into an endocrinology clinic, the doctor is about to prescribe the standard drug propranolol. At the instant the prescription is about to be confirmed, the system pops up an alert: the patient has a history of asthma; propranolol is a non-selective beta blocker that also blocks beta-2 receptors, so it is not recommended for asthma patients — metoprolol is suggested instead.

This is a story that happened in a real clinical setting, and the alert came from an AI avatar of a respiratory medicine specialist.

In their continued evolution, avatar agents not only help doctors pass on experience among themselves and accelerate the improvement of doctors' clinical skills, but also act as a hub for collaboration, conveying information, aligning judgements and reducing friction among doctors, patients, departments and hospitals — making collaboration something that naturally happens in everyday care.

Following the launch of the "Tairex Agent Hospital" online virtual clinic, Tairex will hold the "Shared Life · Shared Intelligence · Shared Future" Tairex Agent Hospital Multi-Agent Collaborative Ecosystem Launch Event on October 21, 2026, formally releasing a "multidisciplinary human-AI consultation" feature to provide patients with multidisciplinary consultations that intervene earlier, apply to more scenarios, reach a wider audience, cost less, deliver higher-quality decisions and offer stronger closed-loop feedback.

In the future, multiple avatar agents will work together on the same operating system, and their application space in healthcare may be further opened up.

Between doctors and AI, is it collaboration or replacement?

Doctors' sense of the efficiency gains brought by Agent Hospital is immediate.

A neurosurgeon said Agent Hospital is a great help in organizing medical record information and sorting clinical information; once agent opinions are introduced into complex preoperative discussions, it can basically save about half the workload, because no dedicated staff are needed to take notes and omissions in the whole process are reduced. He also acknowledged frankly that agents can currently reach the level of a resident physician in most scenarios, but that there are no relevant content modules yet for discussions of surgical approach or postoperative management.

Overall, doctors remain cautious in their judgement of AI decision support. An internist believes that AI's help is currently mostly limited to a supplementary stage, and that the final decision still has to be made by the doctor based on his or her own experience and with reference to the supplementary information AI provides. After comparing various AI models, this doctor thinks a genuinely useful model must both provide detailed diagnostic opinions based on a patient's brief description and, through ongoing conversation with the patient, offer more directional and accurate diagnostic suggestions.

"This process matters. If it is just a single conversation that provides a large amount of information, it (the agent) will also flag rare diseases with a very low probability, which in practice are rarely encountered. As the communication deepens, the leaning becomes clearer and the conclusion more focused," this internist said, adding that the process resembles a doctor with little experience gradually growing into a senior one.

This analogy is thought-provoking: in doctors' eyes, AI is more like an assistant that is still growing — it mostly broadens the view, fills in blind spots and gives a gentle nudge where the doctor might not have thought of something. The final judgement still has to be made by that human being of flesh and blood, with experience and intuition.

Emergency trauma physicians, meanwhile, see AI's value from another angle.

He described an easily overlooked scenario: many primary-level hospitals have no doctors on duty at night to issue reports; after a car accident at night, a patient who has had a CT scan can only get the images and must wait until the doctor comes on duty the next day to get the report. If there were a CT report agent that produced a report three to five minutes after the images come out, the emergency physician would no longer have to wait and could proceed immediately to the next step of medical treatment or surgery.

"The agents Tairex is building are all about solving the problem of uneven distribution of medical resources, and about enabling patients to receive the most effective treatment in the shortest possible time," the emergency trauma physician said.

On the patient side, attitudes are more varied. In their contact with patient groups, a neurosurgeon noted that patients are generally rather averse to the use of AI for diagnosis and treatment and to AI directly intervening in decisions about treatment plans, but showed no obvious reaction when it came to medical record-keeping. An internist also observed that most patients are fairly accepting of using AI themselves, but when doctors use AI, patients worry that the doctor's level or experience may be inadequate.

This subtle psychological gap also reflects the possible focus of the dispute, which often lies not in the technology itself but outside it.

Industry-level problems, moreover, are thornier than individual impressions. At present, the industry's application of AI in healthcare is concentrated in three directions: AI-assisted drug discovery, health consultation and AI imaging. Yet many serious medical scenarios have not yet received adequate support or been addressed; AI imaging has entered medical insurance without generating incremental billing, and doctors' private use of AI tools struggles to become a scaled application.

Collaboration may be the outcome all parties would prefer.

Tairex's choice is to position AI as a doctor's avatar rather than a doctor's replacement. Avatars can run concurrently, evolve and pass on experience. But the final decision is still made by a human.

In fact, collaboration has a highly corresponding form in clinical care — the multidisciplinary consultation. For the same patient, experts from several different departments discuss the condition together and formulate a treatment plan. For example, for a patient with a complex tumour, the attending department initiates an MDT, convening experts from oncology, surgery, radiotherapy, radiology, pathology and other relevant departments to read the imaging together, analyse the test results and jointly decide on the most appropriate treatment plan.

Yet every doctor and every hospital has a capacity ceiling. Patients often only consider initiating an MDT after having registered in several departments without solving their problem, by which point the optimal treatment window may already have been missed. In addition, MDT applies to limited scenarios, reaches a narrow audience and costs a great deal.

Tairex believes that to solve this problem, AI agents should be used to amplify and pass on doctors' experience, rather than simply trying to build a more powerful AI doctor.

Medicine is an industry of theory plus practice, and the time cost of accumulating experience through clinical work is extremely high. A medical student specializing in rheumatology and immunology knows this well. "Five years of undergraduate study plus three years of a master's, plus three years of a doctorate, is at least 11 years; if you do an academic master's and then an academic doctorate, followed by standardized residency training, it can take 12 or 13 years."

With Agent Hospital, however, medical students can first use the virtual clinic module to work through a few cases before entering a department, so as to understand that department's common diseases. When they return to clinical work, since most patients still have common diseases, the students already know roughly how those diseases behave and what their clinical manifestations are, and they get up to speed much faster.

"It is like a kind of clinical simulation: the patient is not as rigid as a standardized patient, and has a clinical background. In clinical practice, the difference between some real patients and those in the virtual clinic will not be great," the medical student said, adding that this in effect builds a bridge between theory and clinical practice for medical students.

A medical student with a traditional Chinese medicine background said that AI's addition may offer some new lines of thinking and, building on the wisdom of predecessors, more innovation. "Based on avatar agents, one can learn the syndrome-differentiation thinking and prescription thinking of national masters of Chinese medicine, which helps young TCM physicians learn and better pass on this heritage."

If an AI agent has already been trained into a specialist avatar, it can greatly save the time consumed by multidisciplinary consultations.

"Because at that point its role is not to make a specific decision but to offer suggestions; as long as it has sufficient specialist competence to identify the specialist issues in the disease, that is actually enough," an internist said, adding that this is of practical help to doctors' growth.

The doctor further noted that what future medical AI needs to do is follow the whole process, becoming more like a senior, seasoned doctor in order to help junior doctors grow and accumulate experience faster. In this way, by drawing on avatar agents, doctors can gain the experience they need from a small number of patients plus summarising and learning. "This is a more efficient method."

In the view of the emergency trauma physician, some of the agents already deployed both lower the barriers to medical care and level out some of the problems caused by unequal medical resources.

This is exactly what Tairex is doing.

Tairex's build-out is extending from the virtual world to real clinics.

On April 10, 2026, Tairex officially released the "Tairex Agent Hospital" online virtual clinic, opening test applications to licensed physicians and current medical students worldwide. The platform is the main entry point through which the virtual medical world is opened to users, and its core mechanism is to support doctor-side users in building and training their own personal doctor avatar agents.

After doctor users choose a specialty, the system randomly pushes matching virtual patients, and users must complete the full-process simulated consultation of taking a history, issuing examination orders, making a diagnosis and giving treatment recommendations. The platform can reproduce real-world processes with high fidelity and automatically generate and return examination and lab test reports and medical imaging reports.

After the consultation ends, the system shows the distribution of decisions made by multiple doctors for the same virtual patient, and operational errors will not harm the virtual system or the user's real professional reputation, providing a low-risk practice environment for junior doctors and medical students.

The Tairex Agent Hospital virtual clinic currently supports 27 clinical departments, covering the great majority of common diseases, major diseases and rare diseases. During the testing of the Tairex Agent Hospital system at physical hospitals in Beijing, Guangxi and Wuxi, the evolution of high-level doctor avatar agents was preliminarily completed. These AI doctor agents and doctor avatar agents form the initial expert pool of the Tairex Agent Hospital agent plaza.

Purely data-driven AI cannot fully adapt to real clinical scenarios. For example, when a patient in his forties has a cough, a purely AI system will leave no stone unturned, eliminating every possible hidden risk before giving a final judgement. That approach is right in the data space, but in the real world it is overtreatment.

AI is now entering real society, learning, giving feedback and improving in a highly dynamic environment.

It should be noted that AI doctor agents are built purely from algorithms, with no real doctor corresponding to them; their growth takes place in the virtual world, interacting with AI patient agents and acquiring medical knowledge and skills through high-throughput evolution. Doctor avatar agents are different: they correspond to a real doctor and grow under that doctor's "teaching by example".

The approach of the Tairex Agent Hospital system is this: a doctor avatar agent starts from an AI doctor agent as its initialization point, works alongside human doctors in the real world and interacts with human patients, achieving superalignment between the doctor avatar agent and the human doctor through symbiotic evolution.

In this process, the doctor avatar agent learns not only medical knowledge but also the doctor's diagnostic habits, including questioning logic, decision logic and individual preferences. For example, if a patient says they have a cough, it follows up by asking whether there is phlegm; if the patient says yes, it then asks whether it is a dry cough, white phlegm or blood-streaked phlegm.

From training in the virtual world, to deployment in real clinics, to the release of the multidisciplinary human-AI consultation feature, Tairex's path is gradually becoming clear.

In real social space, AI and humans work together, and AI gradually aligns with people's values and diagnostic logic. In the future, multi-agent collaboration can give every doctor a superteam.

At the "Shared Life · Shared Intelligence · Shared Future" Tairex Agent Hospital Multi-Agent Collaborative Ecosystem Launch Event, a number of guests from healthcare, AI and industry will gather to discuss: when every doctor has a superteam, what will real clinical scenarios and the healthcare industry become?