Agentic AI: Adapting Large Language Models to Educate Residents on Time-of-Day Energy Pricing in Fort Collins, Colorado.
Category: Research Poster
Author(s): Megan Kelly, Sierra Nordwald
Presenter(s): Megan Kelly
Mentors(s): Timothy Hansen
Many residents of Fort Collins, Colorado remain unaware of the time-of-day pricing model utilized by Fort Collins Utility, leaving them undereducated on how peak and off-peak electricity rates impact their monthly bills. To address this gap, we propose an agentic AI system designed to provide residents with accurate, accessible guidance on the most cost-effective times to operate major household appliances. This study investigates the extent to which a Large Language Model (LLM) based AI agent can accurately and effectively educate household residents about the operational costs of major home appliances, with the goal of reducing both energy consumption and unnecessary expenses. Our research centers on the adaptation of Large Language Models through fine-tuning and Retrieval-Augmented Generation (RAG), trained on large data sets sourced directly from Fort Collins Utility alongside question-based data sets developed to support model evaluation. We anticipate that an AI agent built using these combined approaches will outperform a baseline LLM in improving accuracy to assist residents' in understanding the operational costs of running in home appliances. Resulting in a measurable increase in energy cost awareness and a reduction of unnecessary electricity expenses. Ultimately, this work aims to produce an AI chatbot capable of serving as a personal assistant embedded in smart home devices. Bridging the gap between educational barriers and the unnecessary costs that are associated with the Fort Collins time-of-day pricing structure.