[2604.00046] Large Language Models for Analyzing Enterprise Architecture Debt in Unstructured Documentation

Computer Science > Software Engineering arXiv:2604.00046 (cs) [Submitted on 29 Mar 2026] Title:Large Language Models for Analyzing Enterprise Architecture Debt in Unstructured Documentation Authors:Christin Pagels, Simon Hacks, Rob Henk Bemthuis View a PDF of the paper titled Large Language Models for Analyzing Enterprise Architecture Debt in Unstructured Documentation, by Christin Pagels and 1 other authors View PDF HTML (experimental) Abstract:Enterprise Architecture Debt (EA Debt) arises from suboptimal design decisions and misaligned components that can degrade an organization's IT landscape over time. Early indicators, Enterprise Architecture Smells (EA Smells), are currently mainly detected manually or only from structured artifacts, leaving much unstructured documentation under-analyzed. This study proposes an approach using a large language model (LLM) to identify and quantify EA Debt in unstructured architectural documentation. Following a design science research approach, we design and evaluate an LLM-based prototype for automated EA Smell detection. The artifact ingests unstructured documents (e.g., process descriptions, strategy papers), applies fine-tuned detection models, and outputs identified smells. We evaluate the prototype through a case study using synthetic yet realistic business documents, benchmarking against a custom GPT-based model. Results show that LLMs can detect multiple predefined EA Smells in unstructured text, with the benchmark model achiev...