Getting Data Governance for Regulatory Submissions Right Before AI Gets it Wrong with Cary Smithson

What It Covers

Cary Smithson of Leap Ahead Solutions outlines how life science companies must establish structured data governance frameworks to meet regulatory submission standards like IDMP, PQCMC, and HL7 FHIR, while building the data foundation required for safe, compliant AI adoption across R&D, regulatory, and quality functions.

Key Questions Answered

• •Scope-First Governance: Avoid attempting to govern all data simultaneously. Identify the highest-pain critical data elements (CDEs) — such as product master, substance, dosage form, and manufacturing site data — within a narrow scope first. Demonstrate measurable value quickly to secure ongoing budget, then scale systematically across clinical, regulatory, quality, and manufacturing domains.
• •Business-Owned Stewardship Model: Assign data ownership to business domain experts, not IT. Establish a formal data governance council with named stewards per domain who hold accountability for data exchange, field changes, standards updates, and approval workflows. Tie steward compliance to performance metrics like submission cycle time and right-first-time rates to drive adoption.
• •AI Reliability Depends on Governed Inputs: AI models trained on nonstandard or low-quality data produce erroneous, noncompliant, or unexplainable outputs. A midsize biopharma implementing IDMP-aligned governance achieved 30–50% faster affiliate submissions, 25% fewer health authority queries, and successfully deployed AI-assisted CMC authoring — all within twelve months of launching their data governance program.
• •Regulatory Framework Crosswalk: Maintain a living crosswalk mapping enterprise master data to regulatory data models including RIM and eCTD, with full traceability from source systems to submission artifacts. For IDMP and SPORE, harmonize product and substance attributes across systems. For PQCMC, build CMC data models reflecting manufacturing process parameters, control strategies, and analytical methods fed directly from governed sources.
• •Governance as Persistent Program, Not Project: Embed data governance into existing SOPs, change control processes, system validation (SDLC), and training workflows rather than treating it as a standalone initiative. Monitor health via KPIs including data quality scores, lineage completeness, issue remediation SLA adherence, and submission right-first-time rates, with a continuous improvement loop managed by the governance council.