Intersectoral Analysis of Enterprise Architecture Documentation Using Large Language Models

Enterprise Architecture (EA) documents capture strategic IT plans of organizations and may exhibit patterns that vary by industry. This paper explores the use of a large language model (LLM) to automatically analyze a cross-sector corpus of EA and IT strategy documents. We develop an approach using GPT-4 to encode documents into semantic embeddings and cluster them to reveal shared themes and sector-specific divergences. Our preliminary results indicate that the LLM can surface industry-specific concerns – for example, banking and insurance documents emphasize regulatory compliance and security, whereas manufacturing focuses on product innovation and agility. Clustering the LLM-derived embeddings grouped organizations by these thematic similarities, confirming the hypothesis. Visualizations such as similarity maps and word clouds are presented to illustrate the distinct clusters. This study demonstrates a novel method for comparative analysis of enter- prise architecture strategies across industries using generative AI, with implications for cross-industry knowledge transfer and benchmarking.