In this article, co-authors Vicente Cruz Mínguez, Head of Data and Advanced Analytics at Cepsa Química, and Marcos Fernández Díaz, Senior Data Scientist at Keepler, discuss the implementation of generative Artificial Intelligence (AI) as a transformative force in business industries. They key focus is on the how it impacts the energy sector, specifically, Cepsa Química and partner Keepler’s application of a generative AI assistant to improve the efficiency of their product stewardship team when responding to compliance queries related to the chemical products they market.
The generative AI assistant uses Amazon Bedrock, a managed service, that offers a choice of high-performing foundation models (FMs) from leading AI companies like AI21 Labs, Anthropic, Cohere, Meta, Stability AI, and Amazon. It offers functions such as security, privacy, safety, and the ability to build generative AI applications.
One of the areas that generative AI has demonstrated its value is in streamlining operational processes, which reduces costs and enhances overall productivity levels. The Safety, Sustainability & Energy Transition team at Cepsa Química, is responsible for all human health, safety, and environmental aspects related to the products manufactured by the company and its associated raw materials. This includes managing a large collection of regulatory compliance documents. Their role includes determining which regulations apply to each specific product in the company’s portfolio, compiling a list of all applicable regulations for a given product, and supporting other internal teams that might have questions related to these products and regulations.
The implementation of the generative AI propels a faster resolution of compliance queries. This is a combination of structured and unstructured data culled from product catalogues and regulatory documents. The Large language models (LLMs) were used to extract information from online sources, and a Retrieval Augmented Generation (RAG) approach used to offer accurate, up-to-date data that dynamically adapts to changes in the knowledge base.
The solution method used includes four main functional blocks: input processing, embeddings generation, LLM chain service, and user interface. They developed two independent modules: one to batch process input documents and another to run interference and answer user queries. The modules performed key tasks such as reading and extracting raw documents and product catalogs, data cleanup and post processing as well as storing extracted embeddings.
The use of the solution delivered quantifiable improvements such as query times, answer quality and operational efficiency. The query times demonstrated how the search times of both senior and junior users had been significantly reduced. Answer quality was improved with the system providing additional context and references; while operational efficiency was straightened as the regulatory query process was accelerated.
In conclusion, the generative AI assistant implementation by Cepsa and Keepler has enhanced efficiency and accuracy in handling compliance queries. The authors then motivate for businesses to commence using generative AI, and provided links to Amazon AWS where business can get started.
