Toward Autonomous Finance: A Multi-Agent System Architecture for the Self-Driving Financial Close
DOI:
https://doi.org/10.15662/qr6hkh58Keywords:
autonomous finance, financial close automation, multi-agent systems, continuous accounting, SAP S/4HANA, explainable AI, SOX governanceAbstract
The financial close remains one of the most labor-intensive recurring exercises in corporate finance. Despite two decades of ERP consolidation, robotic process automation, and continuous-accounting practice, exceptions and judgment-bound steps still keep the controller on the critical path. This paper presents a multi- agent system architecture for an autonomous financial close built on SAP S/4HANA. Five specialized agents handle accruals, intercompany transactions, reconciliations, flux analysis, and disclosures; they act on the ledger only through a governed integration layer, are sequenced by a planning engine over a shared task graph, draw on persistent memory of prior closes, and are bounded by a tool registry that enforces least-privilege authority. Audit logging, segregation-of-duty controls, explainability dashboards, and a human-in-the-loop controllership console complete the design. The architecture was evaluated in a controlled, simulated S/4HANA environment processing roughly 1.2 million transactions across several reporting periods. Relative to the pilot's pre- automation baseline, the close cycle fell from 9 days to 2.8 days, the automated reconciliation rate rose from 52% to 94%, exception resolution time dropped from 16 hours to 3 hours, and manual controller workload declined by 65.9%, while journal posting accuracy reached 99.2% and every agent action was captured in a complete audit trail. The results indicate that autonomous agents shorten and de-risk the close only when they are embedded in explicit governance and auditability frameworks, and that the controller's role shifts accordingly from operator to supervisor
References
[1] A. Kumar and M. Kumar, "Artificial intelligence-driven real-time financial reconciliation in modern ERP ecosystems," Int. J. Adv. Res. Comput. Sci. Technol., Dec. 2024, doi: 10.15662/IJARCST.2024.0706013.
[2] H. Issa, T. Sun, and M. A. Vasarhelyi, "Research ideas for artificial intelligence in auditing: The formalization of audit and workforce supplementation," J. Emerg. Technol. Account., vol. 13, no. 2, pp. 1– 20, 2016, doi: 10.2308/jeta-10511.
[3] K. C. Moffitt, A. M. Rozario, and M. A. Vasarhelyi, "Robotic process automation for auditing," J. Emerg. Technol. Account., vol. 15, no. 1, pp. 1–10, 2018, doi: 10.2308/jeta-10589.
[4] M. Schreyer, T. Sattarov, D. Borth, A. Dengel, and B. Reimer, "Detection of anomalies in large scale accounting data using deep autoencoder networks," in Proc. ACM KDD Workshop Anomaly Detection Finance, Aug. 2017, arXiv: 1709.05254.
[5] P. Pokala, "Artificial intelligence in SAP S/4HANA: Transforming enterprise resource planning through intelligent automation," Int. J. Sci. Res. Comput. Sci. Eng. Inf. Technol., vol. 10, no. 6, pp. 191– 201, Nov. 2024, doi: 10.32628/cseit24106169.
[6] W. M. P. van der Aalst, Process Mining: Data Science in Action, 2nd ed. Berlin, Germany: Springer, 2016, doi: 10.1007/978-3-662-49851-4.
[7] V. Bereznyi, "The role of cloud technologies in the organization of continuous accounting," Econ. Manage. Adm., no. 2(108), pp. 78–83, Aug. 2024, doi: 10.26642/ema-2024-2(108)-78-83.
[8] C. Manda, "Scalable multi-agent architecture for enterprise customer experience: Design patterns and implementation," Int. J. Comput. Eng. Technol., vol. 15, no. 6, pp. 1887–1898, Dec. 2024, doi: 10.34218/ijcet_15_06_161.
[9] L. Kirchdorfer, R. Blümel, T. Kampik, H. Van Der Aa, and H. Stuckenschmidt, "AgentSimulator: An agent-based approach for data-driven business process simulation," in Proc. 6th Int. Conf. Process Mining (ICPM), Kgs. Lyngby, Denmark, Sep. 2024, pp. 97–104, doi: 10.1109/icpm63005.2024.10680660.
[10] L. Cheng, H. He, Y. Gu, Q. Liu, Z. Zhao, and F. Fang, "MARS: Multi-agent deep reinforcement learning for real-time workflow scheduling in hybrid clouds with privacy protection," in Proc. IEEE 30th Int. Conf. Parallel Distrib. Syst. (ICPADS), Oct. 2024, pp. 657–666, doi: 10.1109/icpads63350.2024.00091.
[11] S. Mangalampalli et al., "Multi-objective prioritized workflow scheduling using deep reinforcement based learning in cloud computing," IEEE Access, vol. 12, pp. 5373–5392, Jan. 2024, doi: 10.1109/access.2024.3350741.
[12] Z. Zhu, G. Zhang, M. Li, and X. Liu, "Evolutionary multi-objective workflow scheduling in cloud," IEEE Trans. Parallel Distrib. Syst., vol. 27, no. 5, pp. 1344– 1357, May 2016, doi: 10.1109/TPDS.2015.2446459.
[13] A. Amato, A. Morelli, M. Fogli, R. Galliera, and N. Suri, "Multi-agent reinforcement learning for distributed workflow orchestration at the tactical edge," in Proc. IEEE Mil. Commun. Conf. (MILCOM), Oct. 2024, pp. 64–69, doi: 10.1109/milcom61039.2024.10773787.
[14] J. Li, R. Qin, S. Guan, X. Xue, P. Zhu, and F.-Y. Wang, "Digital CEOs in digital enterprises: Automating, augmenting, and parallel in Metaverse/CPSS/TAOs," IEEE/CAA J. Autom. Sinica, vol. 11, no. 4, pp. 820–823, Mar. 2024, doi: 10.1109/jas.2024.124347.
[15] A. Jayanetti, S. Halgamuge, and R. Buyya, "Multi-agent deep reinforcement learning framework for renewable energy-aware workflow scheduling on distributed cloud data centers," IEEE Trans. Parallel Distrib. Syst., vol. 35, no. 4, pp. 604–615, Jan. 2024, doi: 10.1109/tpds.2024.3360448.
[16] J. R. Kuhn and S. G. Sutton, "Continuous auditing in ERP system environments: The current state and future directions," J. Inf. Syst., vol. 24, no. 1, pp. 91–112, 2010, doi: 10.2308/jis.2010.24.1.91.
[17] D. Y. Chan and M. A. Vasarhelyi, "Innovation and practice of continuous auditing," Int. J. Account. Inf. Syst., vol. 12, no. 2, pp. 152–160, Jun. 2011, doi: 10.1016/j.accinf.2011.01.001.
[18] D. Żółtowski, "Emerging ICT in accounting processes automation: Literature review," Procedia Comput. Sci., vol. 246, pp. 4873–4882, Jan. 2024, doi: 10.1016/j.procs.2024.09.353.
[19] S. D. Ramchurn, D. Huynh, and N. R. Jennings, "Trust in multi-agent systems," Knowl. Eng. Rev., vol. 19, no. 1, pp. 1– 25, Mar. 2004, doi: 10.1017/S0269888904000116.
[20] S. M. M. Kumar, "Enhancing journal accounting and month-end closing processes through AI: A comprehensive analysis," Int. J. Sci. Res. (IJSR), vol. 13, no. 3, pp. 1717–1719, Mar. 2024, doi: 10.21275/es24326100738
