Generative Artificial Intelligence (AI) is rapidly emerging as a transformative force, ready to disrupt and reshape businesses of all sizes and industries. Generative AI allows organizations to combine their data with the power of machine learning (ML) algorithms to generate human-like content, streamline processes, and unlock innovation. Like other sectors, the energy industry is being impacted by this paradigm shift, opening opportunities for innovation and efficiency. One area where generative AI is quickly proving its value is in optimizing operational processes, reducing costs, and enhancing overall productivity.
In a recent development, Cepsa Química and Keepler have implemented a generative AI assistant to increase the efficiency of the product management team in responding to queries about the regulatory compliance of the chemicals they market. To accelerate development, they utilized Amazon Bedrock, a fully managed service that offers high-performance model options from leading AI companies through a single API, as well as a broad set of capabilities for building generative AI applications with security, privacy, and safety.
Cepsa Química, a global leader in linear alkylbenzene (LAB) production and second in phenol production, is aligned with Cepsa’s 2030 strategy for decarbonization and sustainability of its processes through the use of renewable raw materials, development of lower-carbon products, and the use of waste as raw materials.
Within the Digitalization, IT, Transformation and Operational Excellence (DITEX) department of Cepsa, they are working on democratizing the use of AI in their business areas to become another lever of value generation. In this context, they identified product management as one of the areas with the highest potential for value creation through generative AI. They partnered with Keepler, a cloud-focused data services company specialized in the design, construction, deployment, and operation of advanced data analytics solutions in public cloud environments for large organizations, in creating the first generative AI solution for one of their corporate teams.
The Safety, Sustainability, and Energy Transition team at Cepsa Química, responsible for all aspects related to human health, safety, and the environment of the products manufactured by the company, was chosen for the initial implementation. This team is responsible for managing a large collection of regulatory compliance documents, a task that consumes a significant percentage of their time.
To address this challenge, they decided to use generative AI techniques to expedite the resolution of compliance queries more quickly. The solution is based on large language models (LLMs), trained with vast amounts of information, and follows an Augmented by Retrieval Generation (RAG) approach, allowing for dynamic adaptability to changes in the knowledge base without the need to retrain the models.
The built solution includes four main functional blocks: input processing, embedding generation, LLM chain service, and user interface, with independent modules for batch processing of input documents and for responding to user queries through inference.
The batch ingestion module performs tasks such as text extraction from PDF documents and generating vectors using Amazon Bedrock, while the inference module transforms user queries into embeddings, retrieves relevant document snippets, and generates contextual responses using LLM models.
During the development of the solution, they faced numerous challenges, such as data preprocessing and document splitting strategy, evaluating different techniques to improve response accuracy. As a result, they were able to enhance the operational efficiency of the team and accelerate the regulatory query process, saving up to 25% of user time.
Cepsa’s DITEX is now working to identify similar use cases in other business areas of Cepsa Química, with the goal of creating a corporate tool that reuses components from this first initiative and generalizes the use of generative AI in business functions.
via: MiMub in Spanish