Generative AI in SAP Document AI for Document Digitization
Overview
SAP Document AI offers comprehensive
document processing for structured, semi-structured and unstructured data
trapped within various document formats, unlocking its true potential for your
business.
By seamlessly integrating and matching this data with your business information, SAP Document AI significantly reduces manual efforts and minimizes errors.
Leveraging the latest advancements in AI
technologies, SAP Document AI provides enhanced accuracy, ensuring high-quality
data for your organization. Our solution empowers informed decision-making
while cutting processing costs, effectively lowering the total cost of
ownership. All of this is delivered through seamless integration within SAP’s
renowned business applications.
SAP Document AI leverages Generative AI to
transform scanned documents (PDF, JPEG, etc.) into structured business data
like invoices, sales orders, payment advice, or any custom template. This
intelligent document processing integrates advanced OCR, multilingual support,
and AI-based learning to automate and enhance data extraction workflows.
Business Challenges
·
Manual Document Processing
·
Unstructured and Multiformat
Data
·
Language and Regional Diversity
·
Lack of Standardization
·
Slow Adaptation to Changes
·
Compliance and Data Accuracy
·
Limited Integration with
Business Systems
·
Scalability Constraints
Key Benefits
-
Automation of manual data entry
processes.
-
Improved accuracy through
continuous learning and adaptation.
-
Scalable solution for
multilingual and multi-format document processing.
-
Enhanced analytics capabilities
with structured data integration.
-
Reduction in operational costs
and processing time.
Extended Capabilities of SAP Document AI
End-to-End Document Processing
Handles both structured and unstructured data from various business documents.
Advanced OCR & Multilingual Support
Supports over 100 languages and 35+ file formats, including handwriting and
barcode detection.
Generative AI Integration
Uses pretrained transformers and large language models for superior document
understanding and minimal training data requirements.
Preconfigured Business Content
Includes templates for invoices, purchase orders, delivery notes, and more.
Process Flow
1. Document Upload Interface
Users upload scanned documents via a
frontend interface. Supported formats include PDF and JPEG.
2. Storage & Preprocessing
Uploaded documents are stored in a Document
Server in unclassified form for further processing.
3. AI-Powered Conversion
SAP Document AI processes the documents
using advanced OCR, multilingual capabilities, and Generative AI to extract and
structure business-relevant data.
4. Learning & Adaptation
The system continuously learns from user
corrections and historical data to improve accuracy and performance.
5. Validation & Correction
Extracted data is validated and corrected
by users or automated rules to ensure accuracy before finalization.
6. Structured Data Output
Final structured data is saved in a
database and displayed via a frontend for analytics and reporting.
Building a SAP Fiori App with RAP and
Document AI Integration
1. Prerequisites
- SAP Business Technology Platform (BTP) account
- SAP S/4HANA system with RAP (Restful ABAP Programming Model) enabled
- SAPUI5 development environment (SAP Web IDE or Business Application Studio)
- Access to SAP Document AI APIs
2. Frontend Setup (SAPUI5)
Create a Fiori app using SAPUI5 with the following views:
- Upload View: Allows users to upload documents
- Validation View: Displays extracted fields and allows corrections
Sample XML for Upload View:
<UploadCollection id="uploadCollection"
uploadUrl="/upload" change="onUploadChange"/>
<Button text="Submit to Document AI"
press="onSubmit"/>
3. Backend Setup (RAP)
Define a RAP service to handle document metadata and validation logic.
Create CDS views and behavior definitions.
Sample CDS View:
@EndUserText.label: 'Document Processor'
define root view entity ZDOCUMENT_AI as select from zdoc_ai {
key DocumentID : UUID;
FileName : string;
Status : string;
Confidence : Decimal(5,2);
}
4. Integration with Document AI APIs
Use HTTP client in ABAP to call Document AI APIs for document submission and
status retrieval.
Sample ABAP Code:
DATA(lo_http_client) = cl_http_client=>create_by_url(
'https://document-ai.sap.com/api/upload' ).
lo_http_client->request->set_header_field( name = 'Authorization' value =
lv_token ).
lo_http_client->request->set_method( 'POST' ).
lo_http_client->send( ).
5. Validation Logic
Once the document is processed, display extracted fields with confidence
levels.
Allow users to correct mappings using dropdowns and save templates.
Sample UI5 Table Row:
<Table>
<columns>
<Column><Text
text="Field"/></Column>
<Column><Text
text="Extracted Value"/></Column>
<Column><Text
text="Confidence"/></Column>
<Column><Text
text="Correction"/></Column>
</columns>
<items>
<ColumnListItem>
<cells>
<Text
text="{fieldName}"/>
<Text
text="{value}"/>
<Text
text="{confidence}%"/>
<Select
items="{dropdownOptions}"
selectedKey="{correctedValue}"/>
</cells>
</ColumnListItem>
</items>
</Table>
Steps followed in POC on Document AI
using Fiori as Frontend.
Step 1: Choose
the jpg or pdf document from computer and submit to Document AI
Step 2: On
submission the document is uploaded. Press refresh to check the processing
status.
Step 3: Once
the document is processed a green tick mark appears. Once the green tick mark
appears, click the document name.
Step 4: Validate
the field mapping for the document. Generative AI find the document type and
maps the field data to the document fields.
Step 5: On
clicking Validate and Save template, a template for the data will be saved
which will be used for future uploaded documents for similar document type.
Step 6: The
template is saved and automatically used next time.
Step 7: Once
validated, document can be saved.
Step 8: Document
gets saved into backend system.
Step 9: The
saved data can be checked in the system tables.
The Fiori app code using RAP is provided in
the attachment.
No comments:
Post a Comment