Healthcare organisations are no longer simply asking whether to use generative AI. A more useful question is whether a general-purpose model can meet the demands of a clinical environment at all. According to Bertalan Meskó, the answer increasingly points towards a separate category: healthcare-specific AI designed around clinical context, workflows, governance and trust.
This distinction matters because a broadly capable conversational model and a dependable clinical tool are not interchangeable. A model may be able to summarise text, draft a response or explain a concept, yet still be poorly suited to the permissions, documentation practices, handovers and safety expectations that shape day-to-day care.
Why healthcare-specific AI is emerging
Meskó’s central observation is that healthcare is beginning to receive AI products built explicitly for its setting rather than generic tools adapted after deployment. The milestones named in current industry discussion—including ChatGPT Health, Claude Health, Microsoft Copilot Health, Ant Afu and ChatGPT for Clinicians—illustrate the direction of travel: vendors are creating healthcare-focused offerings, interfaces or use cases rather than treating medicine as just another prompt category.
The names alone do not establish clinical utility or safety. They do, however, signal that healthcare has requirements that differ from those of general office work. Healthcare AI may need to work with sensitive information, fit established clinical and administrative processes, account for specialised terminology, and support professional judgement without obscuring who remains accountable for a decision.
For leaders, the important implication is strategic. The choice is not merely between adopting AI and delaying adoption. It is increasingly a question of selecting the right level of specialisation for a defined task: a general assistant for low-risk productivity work, a healthcare-tailored tool for a particular workflow, or no automation where the risk and controls are not yet acceptable.
Where general-purpose models can fall short
General-purpose models can be useful for drafting, structuring information and reducing routine administrative effort. But clinical environments add constraints that cannot be solved by a good prompt alone.
- Workflow fit: Clinicians do not work in a blank chat window. They move between records, referrals, results, notes, teams and time-sensitive decisions. An AI tool must fit the work rather than create another destination that staff must maintain.
- Clinical context: Medical language is highly contextual. The meaning of a symptom, test result or recommendation can depend on a patient’s history, setting and care pathway. Outputs require appropriate review and cannot be treated as independent clinical judgement.
- Trust and accountability: Staff and patients need clarity about what a system does, what information it used, where its limitations lie and who is responsible for checking its output.
- Governance: Healthcare organisations must consider privacy, access controls, data handling, validation, auditability and escalation routes before an AI tool is embedded in care or operational processes.
These are not reasons to dismiss AI. They are reasons to evaluate it as a socio-technical change: a combination of technology, people, processes, policy and clinical responsibility.
Questions for a healthcare AI strategy
Organisations assessing medical AI assistants can begin with a narrow, practical use case. Instead of starting with the broad ambition to “use AI,” identify a workflow where the problem, users and review process are clear. Examples might include administrative drafting, documentation support, patient communication preparation or information retrieval for authorised staff.
Leaders can then ask:
- What decision or task is the tool supporting, and what remains the responsibility of the clinician or employee?
- What data will enter the system, and what rules govern access, retention and sharing?
- How will the organisation assess accuracy, reliability and failure modes for this specific use case?
- Does the tool integrate with existing work, or does it add friction and duplicate effort?
- What training, oversight and incident-response process will accompany deployment?
This approach shifts the conversation from impressive demonstrations to implementation discipline. It also helps procurement, digital-health, clinical and governance teams assess vendors using common criteria rather than evaluating claims in isolation.
Bertalan Meskó as a healthcare AI speaker
For healthcare, pharma, medtech and digital-transformation audiences, Meskó’s perspective provides a focused agenda question: how should organisations distinguish between general AI capability and AI that is genuinely ready for healthcare use? That is relevant to leadership meetings where investment priorities are being set, clinical forums considering adoption safeguards, and conferences examining the future of care delivery.
Rather than framing AI as a single product category, this discussion gives audiences a way to examine workflow design, trust, governance and the human roles that remain essential. For event organisers, it can support programmes on clinical AI, responsible innovation, health-system transformation and the changing relationship between technology and care.
To enquire about Bertalan Meskó for a conference, leadership meeting or corporate event, contact Speakers Connect at info@speakersconnect.com.