Supplier Invoice Data Lake & GenAI-Powered Procure2Pay Assistants

The Supplier Invoice Data Lake & GenAI-Powered Procure2Pay Assistant is a modular solution designed to transform raw invoice data into actionable insights using AWS-native services and generative AI. It consists of three integrated components:

1. Supplier Invoice Data Lake Extracts and enhances PDF invoices using GenAI and OCR technologies. Key-value pairs are extracted and stored in a scalable data lake architecture. Entry point is SAP archive with supplier invoices managed by the client SAP instances. In a 1st step the pdf is digitalized and the raw extracted data is made available by the OCR Engine and stored in the Invoice Data Lake. In Step 2 this raw data is enhanced and enriched by multiple GenAI use cases, for sample for Tail-End Analytics or to analyze Tax postings. This enriched Invoice data can be made available in the Enterprise Data Lak or be used on Use Case Level.

AWS Services Used:

• Amazon Bedrock Data Automation – for OCR and structured data extraction
• Amazon Bedrock / Claude – for GenAI-based Data enhancement & enrichment
• Amazon SageMaker – for orchestration
• Amazon OpenSearch - Serverless collections
• Amazon S3 – for secure storage
• AWS Fargate – for containerized SAP archive processing
• AWS Lambda – for serverless execution

2. Tail-End Spend Analytics Enriches invoice data with item-level details (e.g., product codes, quantities, prices, UNSPSC classification) and uploads it to the invoice data lake. Enables procurement teams to optimize catalog usage, negotiate better prices, and reduce tail-end spend.

AWS Services Used:

• Amazon SageMaker – for orchestration
• Amazon Bedrock / Claude– for classification and enrichment for GenAI-driven data augmentation
• Amazon OpenSearch Serverless – Data layer of the invoice data lake, fast search and indexing
• Amazon S3 – for enriched data storage
• Clients Enterprise Analytics Framework – for analytics and querying

3. Invoice Tax Posting Dashboard and Assistant Generating tax posting proposals for supplier invoices using Claude LLM to verify SAP tax postings and supporting identification of wrong tax postings and triggering tax corrections. Especially valuable for countries with complex tax policies (e.g., Argentina), postings in SAP are done manual without the usual available automation. Provides dashboards for local tax teams to validate and correct postings without deep SAP integration requirements.

AWS Services Used:

• Amazon Bedrock/Claude – for LLM-based tax proposal generation
• Amazon SageMaker – for orchestration
• OpenSearch Serverless – Data Layer and querying
• Clients Enterprise Analytics Framework – for dashboarding and insights
• AWS Lambda – for lightweight integration