I’m designing an invoice approval workflow and debating between pure Pega process automation versus incorporating RPA components. The workflow needs to handle invoice ingestion from email, data extraction, validation against purchase orders, multi-level approvals, and posting to our ERP system.
Our ERP has limited API support but a functional web interface. Some vendors send structured data (XML/JSON) while others send PDF invoices requiring OCR. We process about 5000 invoices monthly with peaks during quarter-end.
I’m particularly interested in perspectives on scalability for growing invoice volumes, exception handling when data extraction fails or approvals timeout, and legacy integration strategies since our ERP modernization is 18+ months away. Should I build everything as native Pega process automation, use RPA for specific steps, or create a hybrid approach? What factors should drive this decision?
Use RPA for the ERP interaction since you lack APIs. Native process automation should handle the workflow orchestration, routing, and approvals. RPA bots can be your integration layer for legacy systems. This hybrid approach gives you the best of both worlds - structured workflow management plus flexible screen scraping capabilities.
I’d challenge the RPA-first thinking for legacy integration. Have you explored database-level integration with your ERP? Many legacy systems expose their data layer even when APIs are limited. Direct database reads/writes can be more stable and scalable than screen automation. RPA should be your last resort, not your first choice, especially at 5000 invoices monthly. That volume will stress RPA licensing and maintenance significantly.
Good point Carlos. We do have read access to the ERP database but write operations require going through the application layer for audit trail purposes. What about the OCR and data extraction piece? That seems like an RPA sweet spot for handling unstructured PDF invoices.