CASE STUDY
Scaling Generative AI Impact Across Data Platforms
Client: TC Energy
potentially saved on operational costs
housed documents
At a glance
Challenge
TC Energy, a leading North American energy infrastructure company, identified a critical dynamic of their business: the consistent generation of new information and documentation for efficient access and utilization across various operations. Currently, their information systems house over 5,000,000 documents and many disparate information sources, which will continue to grow. TC Energy was looking for an improved process and set of tooling to make this critical area of memo creation and documentation discovery more streamlined and scalable.
Result
Pariveda helped TC Energy achieve significant operational efficiencies by leveraging tools like Amazon Bedrock and Anthropic Claude Generative AI models to automate technical memo creation and streamline document management processes.
Impact
This transformation will enable TC Energy to potentially save over $500K annually on operational costs, improve information access across various departments, and significantly reduce the time required for critical data analysis and reporting.
Technologies Used
Anthropic Claude 3, Amazon Bedrock, Amazon Glue Data Catalog, Amazon Lambda, Amazon Step Functions, Amazon S3
TC Energy is a leading North American energy infrastructure company headquartered in Calgary, Alberta.
They develop and operate energy infrastructure, including natural gas pipelines, power generation, and gas storage facilities. Their vast network of nearly 60,000 miles of pipelines traverses plains, deserts, and mountains to provide vital energy to consumers across the continent.
The Challenge
Currently, their information systems house over 5,000,000 documents and many disparate information sources, with more information added every day.
The current approach for technical memo creation takes weeks, which is generating an increasingly untenable backlog of reviews. TC Energy was looking for an improved process and set of tooling to make this critical area more streamlined and scalable. Within this document management lifecycle, they pinpointed two specific pain points ripe for innovation and improvement:
- Technical Memo Creation: After an inspection, an engineer typically spends about four weeks creating a technical memo based on detailed data collected. Each memo involves a comprehensive analysis of pipeline conditions, including assessments for internal and external corrosion, geometric deformation, lamination, cracks, and other defects. With an annual backlog of around 120 tech memos, this process generates approximately 19,200 hours of work each year. Streamlining this process is critical to enhance efficiency and share results more quickly.
- Document Management System: TC Energy employs a document management system with over 5 million records reliant on manual metadata tagging with non-uniform definitions. This system poses risks of human error, time-intensive tagging processes, and disparate storage locations, making it difficult to search for critical information. For example, managing contracts involves analyzing terms, renewal dates, obligations, and compliance requirements. This analysis is often hindered by inconsistent metadata, making it challenging to locate and evaluate specific contract details efficiently.
The Result
Pariveda developed a fully serverless, high-performing platform on AWS to leverage Generative AI models with low operational costs, enhancing memo generation and document management efficiency at an enterprise level.
Pariveda leveraged tools such as Amazon Bedrock with Claude V3 LLM models to help TC Energy build two Generative AI processes.
Generative AI can read and summarize data and documents in seconds, and the quality metadata generated will make it easier for legal teams to identify relevant contracts, customer teams to find the most up-to-date term sheets, and operations teams to see what work has been completed on a particular section of the pipeline.
How Pariveda helped TC Energy use Generative AI models to save time and resources:
- Achieved strong results from the model solely through prompt engineering: supplying rigid instructions and suggesting either categorical-like responses or providing examples of 'what good looks like'.
- Initial prototypes used Retrieval-Augmented Generation (RAG) to augment the model's responses. However, the metadata generated for files using this vector database often produced inaccurate answers.
- The simplified approach to use only prompt engineering combined with AWS serverless services to feed specific file information and relevant instructions to the LLM model resulted in a highly performant system with exceptionally low operational costs.
The Impact
With Pariveda’s help, TC Energy developed a platform using Generative AI models to enhance its operational efficiency using data.
The Pariveda team enabled full automation of report content creation and document metadata generation, freeing up employees’ time for specialized tasks. This initiative has streamlined TC Energy’s document management and significantly improved efficiency across the organization.
This transformation will enable TC Energy to potentially save over $500K annually on operational costs, improve information access across various departments, and significantly reduce the time required for critical data analysis and reporting.
Related specialties
Industry
Energy
SERVICE​
Data & Analytics
Case Studies
Explore more success stories
Want to talk?
Looking️ for️ a️ team️ to️ help️ you️ solve️ a️ complex️ problem?️
INSIGHTS
Article
Article
Perspective
