From Meeting Audio to Requirements: A Generative-AI Multi-Agent Pipeline
Resumo
Requirements elicitation from meetings is an expensive, manual, and error-prone activity, in which practitioners must simultaneously participate in the discussion and record it as formal artifacts. This work proposes a multi-agent system, based on generative artificial intelligence, that automates this process end to end, transforming meeting audio into structured requirements documentation. The architecture comprises four specialized agents responsible for automatic audio transcription, persona identification and user story generation following the INVEST criteria, requirements document generation in two selectable formats (the IEEE 830 standard and a client-specific format), and domain diagram synthesis; the agents communicate through a file-based sequential pipeline that provides loose coupling, fault isolation, and entry-point flexibility. The system was validated on a real government software project, using a manually produced official requirements document as ground truth. The quantitative evaluation covers transcription accuracy (WER of 6.3%), system-actor identification (F1 of 0.96) and the INVEST conformance of the generated stories (52.5% on the AMI corpus and 66.7% on the real domain), coverage against the official document (72.7%), and error propagation and cost across the pipeline, processing a nearly 100-minute meeting in about 11 minutes at a cost of US$ 0.72.
Referências
Anthropic. 2026. Claude Model Family. [link]. Accessed: June 2026.
Mohammad Ubaidullah Bokhari and Shams Tabrez Siddiqui. 2011. Metrics for Requirements Engineering and Automated Requirements Tools. In Proceedings of the 5th National Conference on Computing for Nation Development (INDIACom). New Delhi, India, 10–11.
Yangxiao Cai, Ruiyin Li, Peng Liang, Mojtaba Shahin, and Zengyang Li. 2025. Designing LLM-based Multi-Agent Systems for Software Engineering Tasks: Quality Attributes, Design Patterns and Rationale. ACM Transactions on Software Engineering and Methodology (2025). arXiv:2511.08475.
Andrea Delgado et al. 2024. AI-Driven User Story Generation. In Proceedings of the 2024 IEEE International Conference on Software Analysis, Evolution and Reengineering (SANER). IEEE, 1–12. DOI: 10.1109/SANER60148.2024.00000
Fellow.ai. 2025. Fellow: AI Meeting Assistant. [link]. Accessed: June 2026.
Groq, Inc. 2025. Groq API: Fast AI Inference. [link]. Accessed: June 2026.
Sirui Hong, Mingchen Zhuge, Jiaqi Chen, Xiawu Zheng, Yuheng Cheng, Ceyao Zhang, Jinlin Wang, Zili Wang, Steven Ka Shing Yau, Zijuan Lin, Liyang Zhou, Chenyu Ran, Lingfeng Xiao, Chenglin Wu, and Jürgen Schmidhuber. 2024. MetaGPT: Meta Programming for a Multi-Agent Collaborative Framework. In Proceedings of the 12th International Conference on Learning Representations (ICLR 2024).
Max Hort, Fernando Vallecillos-Ruiz, and Leon Moonen. 2025. Large Language Models: A Survey of Surveys. (2025). Preprint. [link].
Jamie. 2025. Jamie: AI Note-Taking Assistant. [link]. Accessed: June 2026.
Dongming Jin, Zhi Jin, Xiaohong Chen, and Chunhui Wang. 2024. MARE: Multi-Agents Collaboration Framework for Requirements Engineering. arXiv preprint arXiv:2405.03256 (2024).
Thomas Kasakowskij and Joerg M. Haake. 2025. T3 Talk2Text: A Model for Near Real-Time Voice Transcription in Virtual Group Meetings. Discover Education 4 (2025), 146. DOI: 10.1007/s44217-025-00568-6
Garm Lucassen, Fabiano Dalpiaz, Jan Martijn E. M. van der Werf, and Sjaak Brinkkemper. 2016. Improving Agile Requirements: The Quality User Story Framework and Tool. Requirements Engineering 21, 3 (2016), 383–403. DOI: 10.1007/s00766-016-0250-x
Qinyu Luo, Yining Ye, Shihao Liang, Zhong Zhang, Yujia Qin, Yaxi Lu, Yesai Wu, Xin Cong, Yankai Lin, Yingli Zhang, Xiaoyin Che, Zhiyuan Liu, and Maosong Sun. 2024. RepoAgent: An LLM-Powered Open-Source Framework for Repository-level Code Documentation Generation. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing: System Demonstrations. Association for Computational Linguistics, 436–464.
Quim Motger, Carlota Catot, and Xavier Franch. 2025. ShaRE your Data! Characterizing Datasets for LLM-based Requirements Engineering. In arXiv preprint arXiv:2510.18787.
Lahbib Naimi, El Mahi Bouziane, Abdeslam Jakimi, Rachid Saadane, and Abdellah Chehri. 2024. Automating Software Documentation: Employing LLMs for Precise Use Case Description. In Procedia Computer Science, Vol. 246. Elsevier, 1346–1354. DOI: 10.1016/j.procs.2024.09.568
OpenNMT. 2024. CTranslate2: Fast Inference Engine for Transformer Models. [link].
Chen Qian, Wei Liu, Hongzhang Liu, Nuo Chen, Yufan Dang, Jiahao Li, Cheng Yang, Weize Chen, Yusheng Su, Xin Cong, Juyuan Xu, Dahai Li, Zhiyuan Liu, and Maosong Sun. 2024. ChatDev: Communicative Agents for Software Development. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (ACL 2024). Association for Computational Linguistics.
Alec Radford, Jong Wook Kim, Tao Xu, et al. 2022. Robust Speech Recognition via Large-Scale Weak Supervision. arXiv preprint arXiv:2212.04356 (2022).
Shuaicai Ren, Hiroyuki Nakagawa, and Tatsuhiro Tsuchiya. 2024. LLM-Based Class Diagram Derivation from User Stories with Chain-of-Thought Promptings. In Proceedings of the 2024 IEEE 48th Annual Computers, Software, and Applications Conference (COMPSAC). IEEE, 1376–1381. DOI: 10.1109/COMPSAC61105.2024.00181
Bill Wake. 2003. INVEST in Good Stories, and SMART Tasks. [link]. Accessed: June 2026.
Xiangqian Xu, Yajie Dou, Liwei Qian, Zhiwei Zhang, Yufeng Ma, and Yuejin Tan. 2023. A Requirement Quality Assessment Method Based on User Stories. Electronics 12, 9 (2023), 2155. DOI: 10.3390/electronics12092155
