Before AI: Why Data Quality Still Determines What Systems Can Do
Interest in artificial intelligence and advanced analytics has accelerated across public-sector health programs. In many cases, these tools are being positioned as solutions to long-standing challenges in monitoring, forecasting, and decision-making. However, experience from implementation work consistently shows that the effectiveness of advanced analytics is constrained by the quality of the underlying data.
When supporting a federal public health contract focused on improving surveillance and analytic reporting, GPHS observed that early enthusiasm for advanced modeling outpaced the readiness of the data environment. Data were drawn from multiple sources with inconsistent definitions, variable completeness, and uneven reporting timelines. While analytic tools could be applied, their outputs were limited by foundational data quality issues that could not be resolved through modeling alone.
Similar patterns have emerged across other applied analytics and monitoring efforts. Investments in advanced tools often assume that data systems are already fit for operational use. In practice, many public-sector data systems were designed for compliance reporting rather than real-time or decision-oriented analysis. As a result, analytic outputs may appear technically sophisticated while remaining operationally fragile.
In emergency response and time-sensitive program settings, perfect data is rarely available. The challenge is not achieving perfection, but understanding the limitations of available data and designing analytic approaches accordingly. Programs that perform well tend to invest early in data governance, validation processes, and shared understanding across teams regarding how data should be interpreted and used.
AI and advanced analytics can add value when applied within these constraints, but they cannot compensate for unclear data ownership, inconsistent reporting practices, or misaligned incentives. Experience across multiple public-sector initiatives demonstrates that data quality establishes the ceiling for what analytic systems can reliably deliver.
Organizations considering AI-enabled approaches benefit from first asking practical questions:
What decisions will the data support?
How reliable is the data at the point of use?
Where are known gaps, and how are they managed?
Addressing these questions strengthens both immediate decision-making and longer-term analytic maturity. In public-sector health programs, data quality remains the prerequisite for meaningful innovation.