Driving operational excellence with GenAI on AWS
The client, a major UK-based gas distribution network, delivers natural gas to over 11 million homes and businesses through more than 131,000 km of underground pipelines. Operating across key regions, including London, the West and the East of England, it plays a vital role in maintaining national energy infrastructure. With a strong focus on safety, engineering excellence and sustainability, the organization supports the UK’s transition to net zero through initiatives like hydrogen blending and biomethane integration.
The challenge
The organization encountered several operational and technical challenges in managing its engineering documentation landscape. The absence of intelligent automation and scalable infrastructure led to inefficiencies, increased complexity and limited visibility across teams. Key challenges included:
- Manual and inefficient access to engineering procedures stored in unstructured formats like PDFs and DOCX files
- No automation to build or scale the knowledge platform on demand, resulting in delays and bottlenecks
- Lack of semantic search capabilities makes it difficult to retrieve accurate, contextually relevant information
- Inability to track usage or cost attribution, limiting governance and stakeholder accountability
- No elastic infrastructure to support dynamic scaling based on user demand or document volume
- Operational risk due to inconsistent adherence to procedures and reliance on tribal knowledge
These challenges underscore the need for a secure, scalable and AI-powered solution to streamline knowledge access and support engineering operations more effectively.
The objective
The primary objective of this initiative was to modernize how engineering teams access procedural documentation by deploying a scalable, AI-powered solution. The organization aimed to reduce manual effort, improve response accuracy and ensure consistent execution. This included enabling conversational access to documents using GenAI, leveraging AWS-native services for scalability and security and empowering internal teams through effective knowledge transfer.
The solution
To address the challenges of inefficient document access and a lack of automation, a GenAI-powered question-answering system was designed and implemented. The solution leverages a Retrieval-Augmented Generation (RAG) architecture built on AWS native services, enabling secure, scalable and intelligent access to engineering procedures. Key components of the solution include:
- Document ingestion pipeline: Automated extraction and parsing of engineering documents (PDFs, DOCX), with metadata tagging and content chunking for semantic retrieval
- Embedding and indexing: Text vectorization using Bedrock-compatible models, indexed via Amazon Kendra or OpenSearch for fast and accurate search
- RAG pipeline: Combines semantic search with GenAI models to generate source-grounded, contextually relevant answers to user queries
- Web-based interface: A secure, user-friendly portal for engineers to interact with the system, submit queries and provide feedback
- Governance and knowledge transfer: Comprehensive documentation, training sessions and access controls to ensure smooth handover and long-term maintainability
This solution not only streamlines access to critical engineering knowledge but also lays the foundation for future expansion into other document types and use cases.
The impact
The GenAI-powered solution delivered significant operational and strategic benefits by transforming how engineering teams access and interact with procedural documentation:
- Significant improvement in information retrieval speed, helping reduce time-to-decision and accelerate task execution
- Enhanced productivity through faster access to accurate, source-grounded answers, minimizing procedural errors and delays
- Better Total Cost of Ownership (TCO) by leveraging AWS-native services like Amazon Bedrock, Lambda and OpenSearch, eliminating the need for legacy software and infrastructure
- Scalable and elastic architecture, capable of handling growing document volumes and user demand without performance degradation
- Secure and compliant environment, with IAM-based access control and audit logging via CloudTrail
- Empowered internal teams through knowledge transfer and training, enabling independent management and evolution of the solution
AWS services used
- Amazon S3
- AWS Lambda
- Amazon Bedrock
- Amazon Kendra
- Amazon OpenSearch
- AWS IAM
- AWS CloudTrail
- Amazon CloudWatch
- Amazon VPC
