Cepsa Química enhances the effectiveness and precision of product oversight with Amazon Bedrock | AWS Machine Learning Blog

Introduction to Generative AI in the Energy Sector

Generative artificial intelligence (AI) is revolutionizing the way businesses operate, including in the energy sector. By leveraging the power of machine learning algorithms, organizations can streamline processes and unlock innovation. In this article, we explore how Cepsa Química and Keepler have implemented generative AI to enhance the efficiency of their product stewardship team.

Transforming Compliance Queries with Generative AI

Cepsa Química, a world leader in the manufacturing of chemicals, partnered with Keepler to develop a generative AI assistant that revolutionizes how compliance queries related to chemical products are answered. By using Amazon Bedrock and high-performing foundation models, they have significantly reduced search times and improved productivity for the product stewardship team.

Generating Contextual Responses with RAG Approach

To keep pace with rapidly evolving regulatory information, the team adopted a Retrieval Augmented Generation (RAG) approach. This method provides dynamism and adaptability to the knowledge base, allowing for quick responses to compliance queries without the need for constant model retraining. The solution is designed to handle both structured and unstructured data sources effectively.

Developing a Comprehensive AI Solution

The generative AI solution built by Cepsa Química and Keepler consists of batch ingestion and inference modules that work together seamlessly. The system processes documents, retrieves relevant information, and generates contextual responses to user queries swiftly. Knowledge Bases for Amazon Bedrock simplifies the implementation process, aiding in the efficient execution of the workflow.

Overcoming Challenges and Enhancing Accuracy

While developing the system, challenges such as complex document structures and response evaluation were encountered. The team implemented mitigation strategies to ensure accurate retrieval and response generation. An automatic evaluation system was put in place to assess the performance of the models effectively, enhancing the overall accuracy of the generative AI solution.

Improving Efficiency and Productivity

The product stewardship chatbot created through generative AI has transformed how compliance queries are addressed, leading to improvements in efficiency and productivity for the team. By automating responses and providing references to input documents, the solution has empowered users to save time and obtain accurate information promptly.

Future of Generative AI at Cepsa Química

Cepsa Química is exploring opportunities to expand the use of generative AI across various business functions. By identifying similar use cases and leveraging the success of the initial implementation, the company aims to create a corporate-wide tool that maximizes the benefits of generative AI applications. The journey continues towards democratizing the use of AI for generating value across the organization.