Putting AI to work in healthcare

Artificial intelligence is beginning to change healthcare in ways that extend well beyond automated diagnosis. For Elaine Daly, partner and head of business consulting at Forvis Mazars, some of the most valuable applications are already being found in everyday processes that place pressure on clinicians and healthcare systems.

“The most immediate impact is often being felt in the operational and administrative layer of healthcare rather than in headline-grabbing clinical applications. One of the strongest use cases is clinical documentation. AI-powered tools can help generate consultation notes, discharge summaries and referral letters, reducing administrative burden and allowing clinicians to spend more time with patients,” she said.

Medical imaging and diagnostics are also developing rapidly, with AI assisting with X-rays, CT and MRI scans, pathology slides and retinal imaging to identify patterns requiring further expert review.

The emphasis, however, is on augmentation rather than replacement.

“The most successful deployments act as decision-support tools that enhance clinical workflows rather than replacing clinicians,” said Daly. “The consistent lesson is that AI succeeds when it is combined with strong clinical governance, validation and user trust. The most effective implementations are not attempting to replace human expertise. They are helping healthcare professionals make better-informed decisions while improving the efficiency and resilience of the wider healthcare system.”

Beyond diagnostics, AI is being applied to clinical decision support, patient flow, remote monitoring, risk stratification and population-health management. In MedTech, organisations are also embedding AI directly into devices and platforms to support clinicians in real time.

Daly sees particular potential in connecting information traditionally held in separate systems.

“The greatest value comes from connecting data sources that have historically existed in isolation. When clinical, prescribing, demographic, environmental and geographic data can be analysed together, it becomes possible to identify patterns that would otherwise remain hidden,” she said.

AI and advanced analytics can help health systems identify communities at increased risk and improve forecasting of healthcare demand. Real-world evidence is another important application, allowing researchers to understand how treatments perform in everyday clinical practice and identify unmet needs.

But connecting information is easier said than done. “Healthcare environments are typically complex, with electronic health records, laboratory systems, imaging platforms, pharmacy systems and administrative applications often operating independently of one another. Without integration, AI remains confined to isolated pilots rather than delivering value at scale,” said Marc Balbirnie, director and data analytics lead at Forvis Mazars.

Cloud platforms, interoperability standards such as FHIR and HL7, and disciplines including DevOps and MLOps are therefore becoming increasingly important.

Security and data sovereignty add another layer of complexity when highly sensitive patient information is involved. Organisations need to consider not only how data is protected, but where it is stored and processed, particularly when cloud platforms or AI models involve infrastructure outside the EU. The EU AI Act also creates specific requirements for certain high-risk healthcare applications, including some AI-based medical software, AI embedded in regulated medical devices, emergency healthcare patient triage and systems assessing eligibility for essential healthcare services.

Governance is therefore critical, requiring clear accountability, risk assessment, privacy and security controls, model validation and transparent documentation.

“The most effective approach is to build governance into AI initiatives from the outset rather than attempting to retrofit controls later,” said Daly. “As AI adoption increases, organisations must navigate a growing range of requirements, including GDPR, the EU AI Act, sector-specific regulations and, in life sciences, GxP obligations.”

For Daly, responsible AI is not a barrier to innovation. “Well-designed governance creates confidence among boards, regulators, employees and customers. It allows organisations to deploy AI more quickly and at greater scale because the risks are understood and appropriately managed,” she said.

For healthcare and MedTech organisations, the next phase of AI is therefore likely to be less about experimentation and more about integration: putting reliable, governed technology into the hands of professionals where it can improve decisions, reduce administrative pressure and ultimately deliver better patient outcomes.

The full text of the article can be found on the Business Post.