An AI-Driven Copilot Agent Layer for Semantic Business Intelligence using Power BI and Power Automate
DOI:
https://doi.org/10.63282/3050-9416.IJAIBDCMS-V7I3P108Keywords:
Business Intelligence, Large Language Models, Conversational Analytics, DAX Generation, Power BI, Power Automate, Workflow Orchestration, AI Agents, NL2DAX, Enterprise Decision SupportAbstract
Enterprise business intelligence (BI) systems have historically relied on static dashboards and manually authored queries, creating analytical bottlenecks that limit organizational agility. This paper proposes a novel AI-driven Copilot Agent Layer that integrates large language model (LLM) reasoning with Microsoft Power BI semantic datasets and Microsoft Power Automate to enable conversational analytics, automated insight generation, and closed-loop workflow orchestration. The architecture adopts a three-tier design separating the user interaction interface, an intelligent reasoning layer, and the enterprise execution environment. A metadata-aware prompting strategy, combined with few-shot learning and schema-grounded validation, translates natural language queries into syntactically correct and semantically aligned Data Analysis Expressions (DAX). Analytical insights generated by the agent trigger downstream Power Automate workflows, closing the gap between data retrieval and operational action. Empirical evaluation across enterprise-scale sales, finance, and operations datasets demonstrates significant gains in query accuracy, time-to-insight, user accessibility, and decision velocity relative to conventional dashboard-centric workflows. The results establish a scalable, governance-aware foundation for next-generation autonomous enterprise intelligence systems.
References
1. Microsoft, “Power BI documentation,” Microsoft Learn. Accessed: May 28, 2026. [Online]. Available: Microsoft Learn Power BI documentation.
2. Microsoft, “Semantic models in the Power BI service,” Microsoft Learn, Oct. 1, 2025. Accessed: May 28, 2026.
3. Microsoft, “Power BI semantic models,” Microsoft Fabric Documentation, Dec. 5, 2025. Accessed: May 28, 2026.
4. Microsoft, “Datasets — Execute Queries — REST API,” Microsoft Learn. Accessed: May 28, 2026.
5. Microsoft, “Develop with the Execute DAX Queries REST API,” Microsoft Learn. Accessed: May 28, 2026.
6. Microsoft, “Mastering the Execute DAX Queries REST API,” Microsoft Learn, May 7, 2026. Accessed: May 28, 2026.
7. Microsoft, “Microsoft Power Automate documentation,” Microsoft Learn. Accessed: May 28, 2026.
8. Microsoft, “Create and test an approval workflow with Power Automate,” Microsoft Learn, Jun. 30, 2025. Accessed: May 28, 2026.
9. Microsoft, “Standard approvals connector,” Microsoft Learn. Accessed: May 28, 2026.
10. Microsoft, “List of all Power Automate connectors,” Microsoft Learn. Accessed: May 28, 2026.
11. L. Shi, Z. Tang, N. Zhang, X. Zhang, and Z. Yang, “A survey on employing large language models for text-to-SQL tasks,” arXiv preprint arXiv:2407.15186, 2024.
12. X. Zhu, Q. Li, L. Cui, and Y. Liu, “Large language model enhanced text-to-SQL generation: A survey,” arXiv preprint arXiv:2410.06011, 2024.
13. F. Lei et al., “Spider 2.0: Evaluating language models on real-world enterprise text-to-SQL workflows,” arXiv preprint arXiv:2411.07763, 2024.
14. S. Yao, J. Zhao, D. Yu, N. Du, I. Shafran, K. Narasimhan, and Y. Cao, “ReAct: Synergizing reasoning and acting in language models,” in Proc. 11th Int. Conf. Learning Representations (ICLR), 2023.
15. J. Wei, X. Wang, D. Schuurmans, M. Bosma, B. Ichter, F. Xia, E. Chi, Q. Le, and D. Zhou, “Chain-of-thought prompting elicits reasoning in large language models,” arXiv preprint arXiv:2201.11903, 2022.
16. P. Lewis et al., “Retrieval-augmented generation for knowledge-intensive NLP tasks,” in Advances in Neural Information Processing Systems, vol. 33, pp. 9459–9474, 2020.
17. Z. Zhang, A. Zhang, M. Li, and A. Smola, “Automatic chain of thought prompting in large language models,” arXiv preprint arXiv:2210.03493, 2022.
18. P. Jiang, L. Cao, R. Zhu, M. Jiang, Y. Zhang, J. Sun, and J. Han, “RAS: Retrieval-and-structuring for knowledge-intensive LLM generation,” arXiv preprint arXiv:2502.10996, 2025.
