AI will augment, not replace, doctors

Writing for Re:solve Global Health, Xeomics’ Chief Innovation Officer argues that generative AI marks a shift from single-purpose clinical algorithms towards something closer to general medical intelligence — and that the question facing medicine is no longer whether to adopt it, but how to do so safely and for the benefit of patients.

The argument

Machine learning has been assisting radiologists, pathologists and drug developers for years, but each system was built for one task and could not carry its knowledge elsewhere. Large language models change that. Because medicine is organised around language — histories, notes, referrals, guidelines, literature — a system that can read, summarise and reason across all of it can support a clinician through an entire workflow rather than one step of it.

Dr Zarkadakis is careful about what that means. In his framing, “AI functions as a cognitive multiplier”: it absorbs the administrative and information-processing load that clinicians consistently say keeps them from patients, and it surfaces evidence that would otherwise stay buried. The best results in radiology come from clinician and AI working together, not from either alone. Clinical judgement — empathy, ethics, context, trust — remains human, and the goal is to strengthen it rather than compete with it.

Why it matters for equitable medicine

The piece is most pointed on governance. The immediate risks are not autonomous machines but biased training data, opaque models, unequal access to compute, and weak validation. Clinical and genomic datasets are still drawn overwhelmingly from Western populations, and without deliberate investment in diverse, shared data and international collaboration, medical AI risks widening global health inequalities rather than closing them.

That is the problem Xeomics exists to work on. The article’s conclusion — that healthcare’s future belongs to clinicians working alongside intelligent systems, on data that represents the populations they serve — is the case for building national genomic programmes and diverse, consented data assets in the first place.

This article was written by Dr George Zarkadakis and published by Re:solve Global Health on September 10, 2026. It is summarised here with a link to the original; the full text is available on the publisher’s site. The opinions expressed are those of the author.