Presentation (CTI)

OBJECTIVES: Global value dossiers (GVDs) are comprehensive, evidence-based documents that synthesise large volumes of evidence. Their development is time- and resource-intensive, and requires integrating complex data into a scientifically accurate, compelling value narrative. This study evaluated the role of generative artificial intelligence (genAI) in GVD development, considering efficiency, accuracy, value narrative strength and alignment with a predefined product strategy.
METHODS: The ‘Disease Background, Burden and Unmet Needs’ section of a GVD for a hypothetical Alzheimer’s disease treatment was developed via separate retrieval augmented generation-based genAI-assisted and manual workstreams in a head-to-head comparison. The manual workstream followed our standard procedures for GVD writing: outline development (stage one) then section draft development (stage two). Senior review of each draft was blinded to ensure objectivity. Key metrics included time spent fact-checking (accuracy) and editing for strategic alignment (quality) when developing the outline and section draft. Analyses of technical accuracy, strategic alignment and value narrative strength were also informed by qualitative insights.
RESULTS: Time to a final GVD outline was 5.7 hours less in the genAI-assisted workstream versus the manual workstream, representing a 62% time saving using genAI. Within stage one, time savings compared with manual development occurred predominantly at initial outline generation stage (83%), with comparable time spent on outline review and refinement. Although human intervention was required in the genAI-assisted workstream to ensure outline completeness, flow and relevance, revisions were not time-intensive. Following these adaptations, the outline developed in the genAI-assisted workstream was of high quality, with good strategic alignment. Results from the second stage of the project will be presented in the poster.
CONCLUSIONS: Use of genAI for outline development provided time savings whilst maintaining quality and strategic alignment in a head-to-head comparison. When used appropriately, genAI can streamline GVD development, but human expertise remains essential to ensure outputs are accurate, comprehensive and compelling.