P03-16 AutoPIF – A platform to optimize cosmetic regulatory documentation: An AI-Enabled Approach

A Product Information File (PIF) is a mandatory regulatory document to ensure a thorough safety assessment is completed before a product is launched, thus meeting regulatory expectations. A comprehensive PIF includes but not limited to product composition, safety assessments, manufacturing specifications, regulatory information, amongst others. With increasing regulatory complexity and pressure to launch product launches, there is a growing need to reduce the time to author high-quality PIFs. AutoPIF™, an innovative and advanced tool designed to optimize and automate the development and management of PIFs, addresses this by leveraging artificial intelligence (AI) to streamline, validate, and optimize PIF creation – ensuring faster, accurate, and compliant submissions. AutoPIF™ combines large language models (LLMs), a regulatory rules engine, and smart document structuring to transform raw technical files into submission-ready PIFs. It extracts and standardizes key data points from safety data sheets, formulation specifications, toxicological profiles, and certificates of analysis, amongst others. It uses automation and AI methodologies to process input data, enabling an automated generation of compliant PIFs that meet regional and client-specific requirements. They include: • NLP integration with the CosIng database for regulatory compliance. • Information extraction for calculation of exposure and margin of safety (MoS) • Assistance in the preparation of content in tables and sections • Record and store the artwork-related images • Automatically capture summaries related to toxicological profiles, claims, PMS, packaging, etc. • Automated versioning for audit readiness Every draft authored by AutoPIF™ is reviewed by qualified subject matter experts who validate and sign off the document (human-in-theloop). The feedback from the quality review process continuously trains the system at the machine-authoring stage. AutoPIF™ has demonstrated an 60% improvement in effort reduction and 95% accuracy in structured content generation, while consistently performing precise exposure and MoS calculations. AutoPIF™ combines the best of human expertise and the latest AI approaches to achieve high-precision efficient automation in the preparation of PIF. Through the optimization of data extraction, organization, and human-driven validation processes, AutoPIF™ facilitates faster and more accurate PIF submissions that minimize operational costs and accelerate product launch timelines, thereby accelerating market penetration and enhancing overall time-to-market efficiency.