Overview
Tungsten Intelligent Document Processing automates the recognition, classification, and extraction of information from documents. Information from PDFs, scans, images, emails, and other document types is converted into structured data that can be used immediately in back-end processes and systems.
Traditional OCR converts images into text. Intelligent Document Processing goes a step further. It understands the type of document being processed, recognizes relevant data, verifies the results, and determines where human review is needed.
In this way, Tungsten bridges the gap between unstructured information and automated business processes.
Benefits
Tungsten reduces manual document processing and makes information available more quickly to back-end processes.
- Less manual work: Automate the reading and entry of data from documents.
- Higher data quality: Recognize, check, and validate information before it is processed further.
- More straight-through processing: Process documents automatically when recognition and checks provide sufficient assurance.
- Wide applicability: Use the same technology for different document types, processes, and business units.
How Tungsten Intelligent Document Processing Works
A document goes through several steps before the information becomes available as reliable data for the downstream process.
Capture
Documents can be received through various channels and in a variety of formats.
For example:
- PDF documents
- Scans and images
- Email and attachments
- Forms
- Invoices
- Orders
- Contracts and other business documents
Tungsten consolidates these documents for further automated processing.
Classification
Before information can be extracted, it must be clear which document is being processed.
Tungsten analyzes content and structure and automatically classifies documents. This allows it to distinguish, for example, between an invoice, order, credit memo, form, or other document type.
The correct processing model can then be applied automatically.
Recognition & Extraction
OCR and AI recognize text and relevant information in the document and convert it into structured data.
In doing so, Tungsten can recognize data such as:
- Supplier or customer
- Invoice and document numbers
- Dates
- Amounts and VAT
- Order references
- Addresses and identification numbers
- Line item details
The technology analyzes not only the text itself but also the context and structure of the document.
Validation
Recognized data is only usable when there is sufficient certainty regarding its quality.
Tungsten can validate data based on recognition confidence, business rules, and available reference data. Fields that have been reliably recognized can be processed automatically, while discrepancies or uncertainties are handled separately.
This ensures a controlled transition from document to reliable process data.
Human in the loop
Not every document can be interpreted fully automatically.
When data is missing, conflicting, or has been recognized with insufficient reliability, an employee can be involved for verification or correction.
Human judgment is thus used specifically for exceptions rather than for every document.
Corrections can also be used to further improve automatic recognition.
Output & Integration
After recognition and validation, documents and structured data are passed on to the underlying process.
This could be, for example:
- ERP system
- Purchase-to-Pay solution
- Order-to-Cash solution
- Document management system
- Workflow platform
- Archive
- API or other integration
are.
Intelligent Document Processing is therefore not an end in itself, but a preparatory layer for further automation.
From OCR to Intelligent Document Processing
Traditional OCR focuses primarily on recognizing characters and text.
Intelligent Document Processing adds context and intelligence to this. The system determines which document is being processed, which information is relevant, how reliable that information is, and what the next step should be.
This shifts document processing from:
image → text
to:
document → structured and validated process data
AI in Document Processing
AI makes it possible to process documents that may vary in layout, wording, and structure.
Instead of relying solely on fixed positions or templates, Tungsten can recognize patterns, classify documents, and interpret data from a variety of layouts.
AI thus primarily supports the steps where variation and interpretation are required. Validation rules and process logic remain important for determining whether the information obtained can actually be processed automatically.
This combines probabilistic recognition and deterministic checks within a single process.
Tungsten in Finance Operations
Within Finance Operations, Intelligent Document Processing is primarily used where financial processes begin with information that is not yet available as structured data.
Purchase to Pay
Recognize and extract data from incoming invoices, orders, order confirmations, and other supplier documents, and make this data available for further processing within the P2P process.
Order to Cash
Process incoming customer documents, orders, remittance information, and other documents that are part of the O2C process.
Financial Control
Make data from reports, statements, and other financial documents digitally available for audits, analyses, and reconciliations.
Cross-Process Document Processing
Use the same document technology outside of Finance for processes that require converting information from documents into reliable, structured data.