Most clinical AI procurement frameworks were built for somewhere else.
If you are a CMIO, head of digital health, or procurement leader at a health system in the Gulf or Hong Kong, you are about to procure a generation of clinical AI. Most vendors selling into this market have never validated their tools in Gulf Arab or East Asian populations. Most frameworks you have were built for NICE or the FDA.
I build the rubrics that close that gap.
I am a UK-trained emergency physician — MRCEM, UCL — with over a decade of senior-grade experience across five major London emergency departments. I work on clinical AI evaluation methodology for health systems in the Gulf and Hong Kong, independent of any vendor.
The work is structured vendor assessment: eight domains scored independently, with automatic red flags and a cross-domain safety interlock that surfaces the deal-killers a composite score averages away. When I ran the Epic Sepsis Model through the instrument on the public record as it stood in 2021, it scored 37 out of 100 — Do Not Proceed, with the interlock firing on population validity and operational performance, not the headline accuracy figure.
If you are evaluating a clinical AI tool, send me the vendor category and the jurisdiction.