Health system leaders are prioritizing artificial intelligence, but Robert Slepin warns that AI deployments will falter without rigorous data discipline. He argues organizations must first trust their own data—improving data quality, provenance, metadata and interoperability—before applying advanced models. Fragmented EHRs, inconsistent coding and weak governance produce biased or unreliable inputs that can undermine model performance, clinician trust and patient safety.
Slepin recommends a practical roadmap: invest in centralized data teams, standardize pipelines, validate datasets, involve clinicians in data governance, and build continuous monitoring and validation to measure real-world impact. Treating data as a strategic asset reduces model risk, helps meet regulatory expectations and increases the chance that AI will deliver measurable clinical and operational benefits rather than costly, ineffective pilots.
