Hexagon has automated the conversion of industrial standard operating procedures (SOPs) into its AcceleratorKMS procedure lifecycle management software using generative AI on AWS, according to an AWS for Industries blog post from October 2025. AcceleratorKMS, part of the Operations and Maintenance product group in Hexagon's Asset Lifecycle Intelligence division, turns operating procedures for energy and manufacturing facilities into structured, mobile-ready content, but a single refinery or plant may need thousands of documents converted. Service teams had used a rule-based Digitizer tool, which was costly to run at scale and struggled with non-standard formats and multiple languages, and earlier attempts to convert documents to images for an LLM lost formatting such as images and hyperlinks.
Working with AWS Partner and Prototyping and Cloud Engineering (PACE) teams, Hexagon built a serverless workflow. Claude 3.7 Sonnet on Amazon Bedrock uses its multimodal capabilities to extract each page into JSON, preserving tables, merged cells and multi-column layouts without fine-tuning. A system prompt defines the JSON schema, the previous page's output is passed in to keep context across pages, and the constant part of the prompt is cached to cut token use. Documents uploaded to Amazon S3 trigger an AWS Step Functions state machine that runs AWS Lambda functions for pre-processing, page-by-page inference and post-processing, with built-in retries, so large batches, such as 100 documents at once, run in the background. Outputs and accuracy metrics are stored in Amazon DynamoDB, the web front end is built with AWS Amplify and Amazon Cognito, and AWS AppSync pushes real-time status updates.
Users review and correct the extracted content side by side with the original before it is finalised, and their edits are used to improve prompts. Because the source text is kept unaltered, there is full traceability from document to output, which matters in regulated industries. The solution achieves above 90% accuracy at both section and statement level, delivers significant cost reductions per document compared with the previous process, and shortens onboarding of new procedures from days to minutes. "This AI-powered approach represents a transformative leap in how we help customers digitize their operational content," said Neil Singh, who leads strategy for AcceleratorKMS at Hexagon.