Hexagon Automates Manufacturing SOP Document Conversion with AWS Generative AI

Hexagon AB partnered with AWS to automate Standard Operating Procedure (SOP) document conversion for industrial customers in Manufacturing and Energy using generative AI. Solution: Hexagon built a scalable serverless workflow on AWS using Amazon Bedrock (Claude Sonnet 3.7) to extract structured data from SOP documents (PDFs, Word with tables, images) with above 90% accuracy. The solution preserves complex formatting like tables and multi-column layouts that simple text extraction misses. Key technical components: Amazon Bedrock + Claude Sonnet 3.7 for document understanding and JSON extraction AWS Step Functions orchestrating Lambda functions for pre-processing, AI inference, and post-processing AWS Amplify + Amazon Cognito for front-end UI Amazon S3 as data lake, DynamoDB for storing outputs and accuracy metrics Human-in-the-loop validation interface for quality control Key metrics: Above 90% accuracy in document conversion Significant cost reduction vs. manual baseline Batch processing up to 100 documents Real-time UI updates via AWS AppSync The solution maintains context across pages by passing previous page output as additional prompt input. Prompt caching is used to reduce Bedrock API token consumption.