By Sam Liscio, PhD, Industry Advisor and NHS Trust Non-Executive Director
AI in healthcare is genuinely high stakes. Every tension that makes artificial intelligence difficult to govern, from accountability and bias to public trust and institutional readiness, exists in health in its most consequential form.
The ‘AI in Health’ conversation we tend to have focuses on the risks of moving too fast: patient safety, algorithmic bias, the accountability gap when something goes wrong. Those concerns are legitimate, but they often crowd out the risk of moving too slowly, and the harm that accumulates in the gap between what AI could do and what we're willing to deploy now.
A 2022 study in BMJ Open found that 16.5% of hospital admissions are caused or complicated by adverse drug reactions, with a projected annual cost to the NHS in England of £2.21 billion.
READ MORE: AI in the Physical World: Prompts are Cheap, Parts Aren’t
Much of that is preventable today with data and AI that already exists in the system. What's missing isn't the patient data or the models, it's the governance architecture, the institutional confidence and the political will to use them.
Like the UK, Canada is grappling with the same dynamic. When Ontario's Auditor General published a critical report on AI scribes earlier this year, Will Falk, whose CSA Group report on generative AI in Canadian healthcare landed the same week, pointed out that the evaluation was based on 2024-era systems that the industry had long since moved beyond.
Using yesterday's evidence to block today's technology is itself a form of governance failure, and one that tends to go unexamined precisely because it looks like caution.
Health systems were built for stability and deliberation, and those instincts have saved countless lives. But they create real drag when the environment is moving faster than the institution can adapt, and the organisations getting AI right are the ones honest enough to weigh the risk of inaction with the same rigour they apply to the risk of action.
READ MORE: Why the best careers keep stretching you
I’ll be exploring those questions at AICON Belfast on 23 September as part of the panel discussion, “AI in Health: What Getting It Right Actually Looks Like,” alongside Mark Avery, Director of Health Informatics at Cambridge University and Health Innovation East, Scott McKenna, Chief Information Officer at Nova Scotia Health & IWK and Eddie O’Neill, Consultant Medical Advisor, Digital Health & Care Northern Ireland, Department of Health.
Healthcare is the sector where AI must get it right, but also get it right quickly. The risks of moving too fast are real: patient safety, algorithmic bias, accountability gaps. But moving too slowly carries its own harms, delayed diagnoses, inequitable access, and research that takes years longer than it should.
This panel brings together senior leaders in health delivery, data and policy to tackle that tension head on. They explore what responsible AI governance looks like in practice, how to maintain public trust without being paralysed by it, and what needs to be built in the next two to three years to convert health data assets into better outcomes for patients.
Read the Summer 2026 edition free online →
Stay connected with NI's tech community:
