SAP Document AI — Enterprise IDP Platform

SAP's unified intelligent document processing (IDP) solution combining OCR, large language models, and pre-trained transformers for enterprise document automation across SAP's business application ecosystem.

Overview

SAP Document AI consolidates the former Document Information Extraction service into a document processing platform deployed on SAP Business Technology Platform (BTP). The platform uses blended AI methods including pre-trained transformers and large language models to minimize training data requirements. Research lineage includes the CharGrid, BERTgrid, and Charmer papers, predating the LLM era. The platform now supports schema-based zero-shot processing to handle complex documents without task-specific retraining.

Q4 2025 marked a strategic inflection. Four capabilities reached general availability: multimodal vision extraction, email attachment processing, multi-step document workflows, and SAP Cloud Transport Management integration. The workflow engine GA is the most architecturally significant. By adding classification, routing, and automated processing pipelines, SAP Document AI moved from a point extraction tool into an orchestration layer for document processing, as SAP's Q4 2025 Business AI release notes confirmed.

Scale figures disclosed through a Hasso Plattner Founders' Award nomination include 30,000+ customers, billions of documents processed, €2.6 billion in estimated annual business value, and 285x growth in SAP BTP usage for custom document automation since 2020. These figures are self-reported and unverified by independent analysts. No Gartner, Everest Group, or IDC assessment of Document AI specifically has been published. The product is natively embedded in 32 business processes across SAP S/4HANA, SAP Business Network, SAP Concur, SAP Fieldglass, SAP SuccessFactors, SAP Customer Experience, and SAP BTP, with dozens more use cases in development.

SAP positions Document AI as an embedded solution within its enterprise application ecosystem rather than a standalone IDP competitor. Integration into 32 business processes means Document AI reaches customers who never evaluated it as a standalone IDP product. This distribution model differs fundamentally from standalone IDP vendors and is structurally difficult for point-solution competitors to replicate.

How SAP Document AI processes documents

The Q4 2025 GA releases define the current processing architecture across four layers, with mobile deployment and LLM orchestration added in early 2026.

Multimodal extraction allows schema administrators to toggle between text-only and combined text-and-image processing. In vision mode, a multimodal model interprets visual elements alongside text, including hazard pictograms, stamps, signatures, logos, diagrams, and labels. Per-schema control allows cost and performance tuning. Text-only IDP systems have a known structural limitation on visually complex compliance documents. The vision mode directly targets Safety Data Sheets and Declarations of Conformity as primary use cases. SAP's per-schema cost controls suggest awareness of vision model pricing and a focus on selective deployment rather than blanket adoption. No throughput benchmarks or accuracy comparisons against competing IDP vendors are provided.

Multi-step document workflows let users define processing pipelines combining extraction, classification, email processing, content-based routing, and automated processing without external integrations or additional tooling. Workflows trigger automatically via inbound channels including email, API, and mobile, or manually via file upload.

Email attachment processing handles attachments alongside or separately from the email body, improving flexibility for email-based document ingestion channels.

Transport management integration with SAP Cloud Transport Management enables export and import of schemas across development, QA, and production instances, addressing enterprise deployment consistency for teams managing multiple environments.

Mobile deployment via SAP Mobile Start v2.4 extends Document AI to iPhone-based field workflows. A community post on the Mobile Start integration describes a side-by-side document view with extraction results for clerk validation, targeting vehicle onboarding, goods receipt, and quality inspection scenarios where paper-to-digital capture is a bottleneck. The one-day implementation scenario emphasizes low-code configuration over custom development.

LLM orchestration via SAP AI Core allows extraction results to be enriched using multiple LLM partners: hyperscaler-hosted models, SAP's Mistral deployment, and Perplexity for web-search-augmented workflows, accessible via BTP contract. This positions Document AI as a component within a broader AI orchestration layer rather than a standalone extraction tool, enabling enrichment workflows (extraction, then LLM reasoning, then system posting) without leaving the SAP ecosystem.

SAP distinguishes Premium AI capabilities (with generative AI) from Classic AI across its portfolio. Classic AI examples include Payment Advice Extraction in FI-Accounts Receivable. Premium AI examples include document classification and entity recognition with generative AI enhancement. This distinction may reflect regulatory caution in finance and healthcare, where generative AI adoption is slower.

Native integration with SuccessFactors, S/4HANA, and SAP ERP provides multilingual OCR across 110+ languages via next-generation generative AI models, with handwriting detection and barcode recognition. No pricing or plan tier information is publicly available for any of the new features, including vision processing. No availability regions are specified.

SAP has signaled reusable tools to support AI agents handling complex document workflows across industries as the next announced capability. No launch date or technical specification has been provided.

Use cases

HR and employee onboarding

The embedded edition within SAP SuccessFactors Onboarding automates extraction of key fields from national ID documents, including ID type, number, and validity dates, and prompts new hires to validate captured data before submission. SAP claims up to 15% acceleration in overall onboarding cycles and up to 30% improvement in validation accuracy. Both figures are SAP benchmark estimates, not independently verified.

Safety and compliance

Vision-enabled extraction processes Safety Data Sheets and Declarations of Conformity with hazard pictogram recognition and visual element identification. These are document types where text-only IDP systems fail structurally. Teams evaluating open-source alternatives for compliance document pipelines may also consider Unstract's no-code LLM platform, which addresses hallucination mitigation for regulated document workflows.

Finance and procurement

Pre-configured templates for invoices, purchase orders, and remittance advice with continuous learning from user corrections. For SAP-connected enterprise workflows, Hypatos also integrates with SAP for financial document automation and offers a point of comparison on straight-through processing rates. Teams building structured extraction pipelines on top of LLMs may also evaluate LangExtract, Google's open-source Python library for grounded structured extraction from unstructured text.

Manufacturing and field logistics

Embedded AI for document processing in S/4HANA Cloud covers quality certificate processing in Public Edition (2508), goods receipt automation in Private Edition (2023 FPS02), and in-house service initiation in Private Edition (2023 FPS03). SAP Field Logistics integrates Business Entity Recognition and Document Extraction for oil and gas industry-specific document automation.

Public sector

The City of Hamburg automatically classified 6 million documents for aid application processing, demonstrating the platform's capacity for high-volume government workflows. Organizations evaluating SAP Document AI for similar government-scale deployments may also compare xSuite, a SAP-certified accounts payable automation provider processing 80+ million documents annually across 60 countries, as a point of reference on SAP-native document throughput at scale.

Technical specifications

Resources

• SAP Document AI Product Page
• SAP Community: Document AI
• SAP Help Documentation
• SAP Tutorials
• Q4 2025 Business AI Release Highlights
• Mobile Start v2.4 Document AI integration
• AI Core and Perplexity enrichment guide

Company information

SAP SE Dietmar-Hopp-Allee 16 69190 Walldorf, Germany Phone: +49 6227 7-47474 Email: info@sap.com Website: https://www.sap.com

We built SAP Document AI to deliver measurable business value at global scale, securely, responsibly, and embedded in everyday processes, demonstrating SAP's ability to operationalize AI at massive scale.

Tobias Weller, Chief Product Owner and Team Lead, SAP Document AI