Transforming Procure-to-Pay with Generative Artificial Intelligence: An Architecture-Grounded Analysis of AI-Powered Invoice Extraction and Non-PO Invoice Processing in SAP Ariba Environments

Gururaj Veershetty

Consulting Manager /Service Delivery Manager

Correspondence: gururaj.veershetty@outlook.com

Abstract

The proliferation of unstructured supplier documents, particularly PDF invoices, remains one of the most persistent operational bottlenecks in enterprise procure-to-pay (P2P) cycles. Manual extraction, inconsistent data quality at intake, and recurring reconciliation exceptions drive significant labor costs, delay payment cycles, and limit the capture of early-payment discounts. This paper presents a rigorous, architecture-grounded analysis of Generative AI (GenAI)-powered invoice PDF extraction integrated with SAP Ariba Buying and Invoicing. Drawing on primary design analysis, benchmark-aligned case evidence, and enterprise integration patterns (REST API, SAP Cloud Integration Gateway), we articulate the functional flow, value levers, implementation prerequisites, and operational risk controls for this capability. A human-in-the-loop validation model and master data governance (MDG) integration are examined as critical guardrails. Complementary P2P capabilities—AI-guided buying, intelligent supplier onboarding, and the SAP Joule generative AI copilot—are situated within a unified enterprise transformation framework. KPI recommendations and a measurement approach grounded in activity-based costing are provided. Findings indicate that organizations achieving sustained catalog coverage and sufficient non-PO invoice volume can realize positive ROI within 12–18 months, with measurable reductions in touchless rate gaps, exception backlog, and per-invoice processing cost.

Keywords: Generative AI | Invoice Processing | SAP Ariba | Procure-to-Pay | Large Language Models | Human-in-the-Loop | Accounts Payable Automation | Master Data Governance