FRoSTA used SAP Build Process Automation and SAP Document AI on SAP BTP to create a largely touchless invoice booking process for 30,000 annual invoices across paper, PDF, and Excel formats.
A trainee with no programming background implemented about 90% of the automation in three months, supported by SAP Discovery Center content, SAP Business Accelerator Hub artifacts, partners, and the Joule copilot.
The new solution now processes about 60% of fully received goods invoices automatically and routes the remaining 40% into rule-based approval or exception handling, significantly reducing cycle times and manual workload in the procure-to-pay process.
FRoSTA has built a strategic partnership with SAP over several years and has a strong digital transformation agenda. The company had already been using SAP Business Technology Platform (SAP BTP) for several years, integrating SAP S/4HANA with SAP SuccessFactors solutions and third-party tools. Expanding on its strategy for SAP BTP, the business developed analytics scenarios leveraging the SAP Datasphere and SAP Analytics Cloud solutions. The accessibility, flexibility, and ease of integration led FRoSTA to look at SAP Build solutions on SAP BTP to help extend and automate its cloud ERP following a clean core approach.
Increasing the automation rate in invoice processing was identified as the first use case within its core processes. The company chose the SAP Build Process Automation and SAP Document AI solutions for their smooth integration with SAP and third-party software. A trainee in the IT and digital business transformation team led in developing the scenario within three months while cooperating closely with the data and AI manager and an expert on SAP BTP from the digital transformation team.
The project used a mission from the SAP Discovery Center site as a starting point. FRoSTA was able to access a public artifact repository of business content on SAP BTP using SAP Business Accelerator Hub. Partner sovanta AG conducted a two-week training session with the food manufacturer to assist with the implementation, and partner Amista provided support for adopting best practices with SAP Build Process Automation. Other than that, FRoSTA’s trainee performed 90% of the work, despite having no programming background, using a search engine and thirdparty
AI tool as well as the Joule copilot for programming.
While the team was happy with the recognition rate of the template-based approach to information extraction, the large variety of invoice formats made training for each template format impractical. After consulting with the product management team at SAP, FRoSTA switched to the premium edition of SAP Document AI. This edition uses a large language model (LLM) for capturing, validating, and processing invoice data and was thus better suited to process multiple document formats.