Fast-track SOP processing using Amazon Bedrock

This article explores approaches to using Amazon Bedrock and generative AI to identify relationships between regulatory changes and Standard Operating Procedures (SOPs) in the healthcare and life sciences industry. The study focused on FDA biologics procedures and tested three main methods:

• Full document matching using Amazon Bedrock's Claude 3 Opus model
• Text similarity approaches including vector embeddings and keyword search
• Taxonomy-based topic matching to map documents to a hierarchical topic structure

Key findings revealed that full text matching and taxonomy-based topic matching were most effective in accurately identifying related SOPs, while text similarity methods were less reliable. The research demonstrates how generative AI can help organizations quickly assess the impact of regulatory changes on their internal procedures.