Mark Kelly
AI Ireland Founder, CCO Alldus
As healthcare demand surges, AI avatars are being deployed across millions of patient interactions — but can they truly ease pressure without compromising care?
Ireland’s workforce crisis is stark. The HSE reported more than 6,500 nursing and midwifery vacancies in 2024, while clinicians often lose up to two hours daily to administrative paperwork. To address these challenges, health systems are increasingly adopting AI as a digital assistant to support clinicians rather than replace them.
How AI is cutting admin time
The scale of experimentation is already significant. In the US, Hippocratic AI’s avatar agents have handled over 115 million clinical interactions, calling patients post-surgery to check vitals, provide medication reminders and answer routine questions. The goal is to free human clinicians for complex cases. Trials with major US hospitals are ongoing, though peer-reviewed results are still awaited.
Meanwhile, Microsoft’s Dragon Copilot targets the documentation burden directly. Using voice recognition and “ambient listening”, it drafts notes and referrals automatically. Simon Wallace, Chief Medical Information Officer at Microsoft UK, reports that over 200 clinicians have tested it, saving up to an hour daily, time redirected from typing to patient care.
Risks associated with generative AI
Yet significant risks remain. Both systems rely on generative AI, which can produce confident but incorrect answers, a serious concern in medicine. Legal liability remains unclear: who’s responsible if an AI misses a symptom? Data protection presents another challenge under GDPR and the new EU AI Act, which classifies medical AI as “high-risk”. Transparency over data use and model accuracy remains limited. Cost could also deepen inequality, with private hospitals adopting AI faster than public systems can afford.
Ireland has made progress in targeted areas. The Mater Hospital’s AI use in radiology has accelerated diagnosis and triage, flagging urgent cases with over 90% accuracy, a model of smart deployment. But Hippocratic AI’s massive US rollout demonstrates that scale doesn’t equal clinical success.
AI should amplify human effort, not replace it. For Irish healthcare, the path forward requires careful judgment. Adoption must be guided by rigorous evaluation and transparency, deployed where evidence supports it, and always grounded in empathy and human connection. Technology can enhance clinical capacity, but only if implemented with the same care we expect clinicians to show their patients.
