Medical AI is unlikely to become a single uniform market with one winner for every task. Large platforms may dominate important distribution channels while specialists remain useful in particular settings, jurisdictions and learning needs. Survival is not guaranteed, however: a specialist must offer something customers still value when a broad platform improves.
This is scenario analysis as of 15 October 2026, not a market-share forecast. It is published by iatroX and includes its clinical-reference and education model among the specialist approaches considered.
Scale helps, but distinguish the mechanism
A large user base can strengthen a brand, make partnerships attractive and spread development costs across more customers. None of those advantages is identical to a network effect.
A network effect means that the service becomes more valuable to a user because other relevant users participate. A clinical-reference tool may benefit from feedback and professional recognition, but a rising registration count does not by itself demonstrate that mechanism. Nor does access to more users automatically permit reuse of their clinical information for training.
The practical investor question is whether scale improves the product, lowers costs, strengthens distribution or merely increases publicity. Those mechanisms can coexist, but they should be assessed separately.
Healthcare is fragmented in ways that matter to the product
The same medical term can sit inside different care pathways. A referral may require different information, treatment availability may differ and the responsible professional may work under different arrangements. A product that is convenient in one setting may need substantial adaptation elsewhere.
There is also fragmentation within a country. A ward nurse, a community pharmacist, a GP trainee and a hospital procurement team do not necessarily need the same interface or evidence. Even when they share clinical knowledge, their tasks and responsibilities differ.
The MHRA's intended-purpose guidance, checked on 15 October 2026, makes the function, intended users and setting relevant to software assessment. It is one reason that "healthcare AI" is too broad a label for comparing products or assuming that one successful deployment generalises everywhere.
A specialist can win a task without owning the whole day
Consider a fictional clinician who uses an institutional assistant for documentation but a separate platform for a postgraduate examination. The first helps complete clinical work. The second organises deliberate practice around an assessment and a timetable.
A broad platform might eventually address both needs well. Until then, the existence of an integrated assistant does not make the examination need disappear. The specialist's opportunity is to provide a coherent learning activity, not simply a second chat box containing similar medical text.
For September 2026, iatroX's examination offering includes adaptive sequencing and spaced repetition, while Tutor works from the attempted question. These are concrete product functions, not evidence that every learner should choose iatroX or that its outcomes exceed those of a larger platform.
Specialisation can also become a trap
A narrow product may depend on a feature that an incumbent can readily bundle. It may serve a small market whose support requirements exceed the revenue available. It may rely on a single integration partner that controls access to customers.
A fictional specialty assistant could have excellent terminology but poor maintenance of local guidance. Another could solve a real problem but require an expensive custom implementation at every site. Being specialised is therefore not sufficient; the service must be useful, maintainable and commercially coherent.
The strongest specialist asks what it does better for the defined task and what evidence supports that claim. The weakest relies on a professional label while reproducing generic model behaviour.
Platforms can incorporate specialists without eliminating them
A large workflow provider does not necessarily need to build every source, model or educational component internally. It may licence content, provide an integration route or partner with a specialist whose expertise is costly to reproduce.
Wolters Kluwer's March 2026 Microsoft announcement provides a concrete example of maintained clinical content entering another company's workflow. HL7 CDS Hooks provides a technical pattern for modular decision-support services. Neither guarantees open commercial terms or effortless interoperability.
In a partnership model, the specialist gains reach but may lose control over the interface and some economics. Its continued value depends on the quality of the component and the terms of the relationship, not simply the existence of a partner logo.
Acquisition is one outcome, not the definition of success
A platform may acquire a specialist to obtain expertise, integrations, customers or a product already suited to a particular setting. Doctolib's May 2026 Medicus announcement illustrates expansion through a company positioned in UK primary-care software.
That transaction does not establish which company will be acquired next or that all independent products should pursue the same route. Some specialists may remain profitable independently; others may become component suppliers or focus on a smaller professional audience.
The strategic test is what remains valuable after integration into the larger organisation. A customer relationship can weaken if the acquired service loses its distinctive purpose, while a carefully preserved specialist capability may become more useful through broader distribution.
Three market structures could coexist
A dominant-platform structure may emerge within a particular institution, where a core system becomes the main interface. A specialist ecosystem may remain around it for evidence, professional learning or unusual workflows. A cross-institutional destination may continue to serve clinicians who value a consistent tool across workplaces.
These structures are not mutually exclusive. The same clinician could encounter all three during a week. The real question is how responsibilities, permissions and commercial relationships work across them.
A universal market-share prediction would obscure that variation. A more useful forecast identifies the task, buyer and jurisdiction, then asks which suppliers can meet the need with sustainable economics and appropriate evidence.
What specialists need to demonstrate now
A specialist should be able to describe the job it serves, the alternative available today and the improvement it offers. It should show how it maintains sources or content, handles errors and supports customers when the task becomes difficult.
It should also make switching or export practical where professional records are involved. Retaining customers through useful work is a stronger proposition than making their learning history or workflow configuration inaccessible.
The iatroX combination of reference and learning is one possible strategy. It must still meet the same standards of transparency and usefulness. A separate interface does not automatically supply independent verification, and a learning catalogue does not establish educational effectiveness.
Verdict by scenario
For a hospital seeking operational consistency, a platform may be the sensible foundation, with specialist additions where they justify the complexity. For a clinician moving between workplaces, an independent professional tool may offer valuable continuity. For a learner with a defined assessment goal, a specialist learning environment may be preferable to a broad assistant.
The likely contest is not one winner against every specialist. It is which parts of healthcare software become sufficiently standard to bundle, and which still reward focused expertise, trust and a better fit to the user's work.
Frequently asked questions
Does a large user count prove a network effect?
No: scale can support brand or cost advantages without making the product inherently more valuable because additional users joined. The mechanism needs to be demonstrated.
Can a specialist survive without owning an EHR?
Yes, through a distinct task, professional relationship or integration role. It still needs a compelling reason for customers or partners to use it.
Will every specialist eventually be acquired?
There is no basis for that prediction. Independent operation, partnerships and component supply are also plausible routes.
Explore focused examination and clinical-learning pathways →