Healthcare systems race to decide how deep AI should go: PwC

Addressing privacy and ethical concerns ranks as the toughest obstacle.

Healthcare organisations are no longer debating whether to adopt artificial intelligence (AI) but how deeply to embed it into their operating models, according to a PwC report. 

AI's value will increasingly be judged by its impact on patient outcomes and system performance rather than the number of tools deployed, it said.

Today, most healthcare AI is confined to discrete, well-bounded tasks, such as transcription tools displacing manual note-taking, whilst AI-supported coding and rostering are cutting administrative bottlenecks and freeing clinician time.

The report expects that focus to shift toward prevention and cross-functional coordination, with AI operating earlier — before the clinical encounter — rather than purely as back-office support.

Instead of responding to patient deterioration once it becomes clinically visible, AI systems are expected to flag risk earlier and prompt intervention before problems escalate.

Machine learning and data analytics are already being applied to children's cancer treatment, predicting relapse likelihood and supporting clinicians in making more informed decisions with greater accuracy.

PwC ties this shift to system-level gains such as fewer avoidable hospital admissions, shorter length of stay and reduced downstream capacity pressure.

However, as generative AI becomes widely accessible, informal and unmanaged use is outpacing organisations' ability to govern it, creating a growing divide between organisations with defined AI governance and those with fragmented, unclear accountability.

Building trust and addressing privacy and ethical concerns rank as the toughest obstacles, followed by workforce education and resistance to change — both now outweighing issues like data quality or outdated infrastructure.

Moreover, PwC's 2025 Global Digital Trust Insights survey found only 24% of healthcare leaders are confident in their compliance with privacy regulations, and just 19% are confident in AI-regulation compliance.