A growing number of healthcare organizations are deploying AI without addressing fundamental data quality issues, according to the data governance director at a major US health system, who warned that automation simply accelerates the impact of bad data rather than fixing it.
The warning comes as hospitals race to implement AI tools for clinical decision support, prior authorization, and revenue cycle management. The core problem is that AI models trained on inconsistent, incomplete, or biased data produce unreliable outputs — and the speed at which AI operates means errors compound faster than with manual processes.
Health systems investing in AI governance frameworks are still the exception rather than the rule, the director said. Most organizations lack standardized data definitions, consistent coding practices, and validation pipelines for their training datasets.
Industry groups have begun developing data quality benchmarks specifically for healthcare AI applications, but widespread adoption remains months or years away.
