JMIR Publications

ABSTRACT

Background:

Healthcare systems face unsustainable cost pressures while artificial intelligence reaches clinical parity across multiple domains. Yet, measurable cost savings remain limited, largely due to implementation challenges rather than technological shortcomings.

Objective:

To develop and apply a Bayesian simulation framework to estimate AI-driven cost savings and implementation risks in UK and Australian healthcare systems.

Methods:

We built a Bayesian simulation model to estimate AI-driven cost savings in radiology, workflow optimization, and workforce optimization within UK and Australian healthcare systems. Effectiveness parameters were drawn from a comprehensive literature review and combined with implementation risk distributions and adoption rates. Monte Carlo simulations (N = 1,000) generated probabilistic cost projections.

Results:

Projected annual savings were $1.14 billion (95% CI: $0.91–1.33 billion) for the UK and $0.71 billion (95% CI: $0.54–0.91 billion) for Australia. Workforce optimization accounted for the largest share (UK 73.2%; Australia 73.8%), followed by workflow optimization and radiology. Implementation risks substantially reduced realized savings, with posterior risk estimates ranging from 42.5% to 68.0% across sectors. Scenario analysis showed potential savings from $0.30–2.49 billion in the UK and $0.15–1.64 billion in Australia, highlighting that organizational factors, rather than technology performance, dominate economic outcomes.

Conclusions:

AI can generate significant healthcare cost savings; however, success relies on effectively managing implementation risks. The Bayesian framework provides policymakers with probabilistic savings estimates to support evidence-based resource allocation, while explicitly accounting for uncertainty in adoption. Clinical Trial: Not Applicable.