AI-Powered Invoice Processing in ERP Systems: A Reference Architecture for Confidence-Governed Touchless Accounts Payable Automation

This paper has argued that the rapidly maturing literature on AI-powered invoice extraction, while substantially advancing the accuracy with which structured data can be pulled from unstructured invoice documents, has left the architecture connecting that extraction to trustworthy, automated ERP posting comparatively underdeveloped. Grounding the problem in information quality theory (Wang and Strong 1996) and control theory (Ivanov et al. 2018), this paper proposed an eight-layer reference architecture extending invoice extraction into a complete invoice-to-pay pipeline, with a confidence-governance layer as its central contribution: a weakest-link aggregation of extraction, validation, matching, and tax and GL determination confidence, routed through calibrated, recalibratable thresholds to touchless posting, human review, or rejection. An illustrative scenario demonstrated the architecture's central design principle concretely, that a single low-confidence dimension appropriately withholds full automation even when other pipeline stages are highly confident. For researchers, the paper's primary contribution is a formal governance mechanism addressing a gap between the mature extraction literature and enterprise financial control requirements; for practitioners, its primary contribution is an explainable, auditable alternative to ad hoc touchless-eligibility configuration. Empirical validation of the propositions advanced in Section 9, through the reference implementation and case studies identified in Section 12, remains the necessary next step before this architecture's practical value can be established with confidence.

Hevner, A. R., March, S. T., Park, J., & Ram, S. (2004). Design science in information systems research. MIS Quarterly, 28(1), 75–105. https://doi.org/10.2307/25148625 Ivanov, D., Sethi, S., Dolgui, A., & Sokolov, B. (2018). A survey on control theory applications to operational systems, supply chain management, and Industry 4.0. Annual Reviews in Control, 46, 134–147. https://doi.org/10.1016/j.arcontrol.2018.10.003 Krieger, F., Drews, P., Funk, B., & Wobbe, T. (2021). Information extraction from invoices: A graph neural network approach for datasets with high layout variety. In F. Ahlemann, R. Schütte, & S. Stieglitz (Eds.), Innovation through information systems: Volume II: A collection of latest research on technology issues (pp. 5–20). Springer. https://doi.org/10.1007/978-3-030-86797-3_1 Liu, X., Gao, F., Zhang, Q., & Zhao, H. (2019). Graph convolution for multimodal information extraction from visually rich documents. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 2 (Industry Papers) (pp. 32–39). Association for Computational Linguistics. https://doi.org/10.18653/v1/N19-2005 Palm, R. B., Laws, F., & Winther, O. (2019). Attend, copy, parse: End-to-end information extraction from documents. In 2019 International Conference on Document Analysis and Recognition (ICDAR) (pp. 329–336). IEEE. https://doi.org/10.1109/ICDAR.2019.00060 Palm, R. B., Winther, O., & Laws, F. (2017). CloudScan—A configuration-free invoice analysis system using recurrent neural networks. In 2017 14th IAPR International Conference on Document Analysis and Recognition (ICDAR) (Vol. 1, pp. 406–413). IEEE. https://doi.org/10.1109/ICDAR.2017.74 Peffers, K., Tuunanen, T., Rothenberger, M. A., & Chatterjee, S. (2007). A design science research methodology for information systems research. Journal of Management Information Systems, 24(3), 45–77. https://doi.org/10.2753/MIS0742-1222240302 Saout, T., Lardeux, F., & Saubion, F. (2024). An overview of data extraction from invoices. IEEE Access, 12, 19872–19886. https://doi.org/10.1109/ACCESS.2024.3360528 Wang, R. Y., & Strong, D. M. (1996). Beyond accuracy: What data quality means to data consumers. Journal of Management Information Systems, 12(4), 5–33. https://doi.org/10.1080/07421222.1996.11518099