Document AI Consumption Models: Embedded versus Custom
Document AI is commercialized in two ways: Embedded AI and Custom AI. Embedded AI is tightly integrated into SAP applications and follows a straightforward model: usage is billed in AI Units, based on requests and users. This makes it easy for customers to adopt out-of-the-box best practices for document processing with complete transparency of consumption. We will take a closer look at the details of this model in the next section.
Custom AI, in contrast, is designed for building fully tailored document workflows and automations. Instead of AI Units, it is billed via BTP Enterprise Agreement credits or PAYG, with consumption measured in blocks of documents and requests.
In short, Embedded AI with AI Units is the primary model for document processing inside SAP applications, while Custom AI offers a flexible alternative for bespoke scenarios.
How Embedded Document AI is Commercialized
Embedded Document AI has two main components that drive consumption: the workspace (per user) and the embedded edition (per request).
Document AI Workspace is licensed per user, typically for administrators who manage authorization, configuration, and design of document workflows. Each user includes 100 GB of storage per month and is priced at 100 AI Units per month.
The Document AI Embedded Edition is tied to document processing itself. Here, each processed artifact — for example, a page classified, translated, split, or extracted — counts as one request. Each request consumes 0.1 AI Units, making costs scale directly with actual document volumes.
Together, these two dimensions ensure that customers pay for both the administrative setup capacity (workspace users) and the actual document processing performed (requests), creating a transparent and predictable model for scaling document automation
Lesson Summary
Document AI is commercialized through two paths: Embedded AI which uses AI Units for both admin workspaces (per user) and document processing (per request), and Custom AI, which is billed via BTP EA or PAYG. This model ensures that customers can either adopt ready-to-go, embedded capabilities with transparent AI Unit pricing or design fully tailored workflows with flexible BTP-based consumption.