For building practitioners, staying current with guidelines and regulations has always been crucial. Traditional methods relies predominantly on researchers reviewing and comparing international building standards and is known to be time consuming. Leveraging recent advancements in large language models (LLMs), which have enabled more efficient information retrieval/comparison techniques on various tasks, we propose a novel framework in this paper that extracts, compares and varies design briefs/inputs for NZEB designs, in particular when new guidelines and requirements arise from new international building codes/requirements. To extract requirements from building codes and standards, we rely on unsupervised Named Entity Recognition (NER) techniques. However, the state-of-the-art approach in extracting relationships established in papers can be challenging, especially before the emergence of more generalized LLMs. To address this challenge, we use kor, a tool that enables the definition of various free text and/or numbered extraction based on description and definition, and few examples that the model can learn from. The instruction is passed as a JSON object to the LLM, which analyzes specific traits from the model following a tailored recipe. For a proof of concept, we will be comparing three versions of ASHRAE Standard 55, using the LLM to assess specific sections related to clothing factor determination within each standard. While larger-scale applications may present challenges in splitting and retrieving the overall standard as a text corpus, our proposed framework offers evidence that systematic information extraction from building standards using LLMs is feasible. By adapting to evolving requirements in real-time, our proposed approach ensures compliance with new standards while providing a scalable solution for building practitioners. The use of LLMs offers a promising path forward in managing complex building codes and standards, enabling NZEB designs in an ever-changing regulatory requirements.