Industry
Public Administration / Social Security
Work Area
Incoming Mail Automation, Document Management & AI-Powered Classification
Project Scope
Automate 20–30% of incoming emails through intelligent pre-screening and AI-powered categorization
Initial Situation
Until now, customer emails were routed to general inboxes at PVA and had to be reviewed entirely manually by staff in the KOR department (service department). Employees had to identify the customer, understand the issue, and forward the email to a technical mailbox. From there, archiving, PDF conversion, and linking to the customer were initiated manually, and the emails were tagged. A particular “pain point” was recurring bulk mail, such as life event notifications, which arrived in enormous quantities within a short period of time and severely bogged down the department.
Objective
- Reducing the workload on the service department (KOR): Significant reduction in the number of customer emails that need to be reviewed manually.
- Automation of “low-hanging fruit”: Fully automated processing and assignment of approximately 20–30% of standard inquiries.
- Intelligent cataloging: Supporting case workers through AI-powered document cleaning and tagging.
Success Factors
- The Pragmatic Tech Mix: Success lies in the intelligent combination of simple, error-free methods (regular expressions/full-text search for insurance numbers) and targeted use of AI (image recognition, classification).
- Focus on the Pain Point: The targeted solution for life insurance confirmations immediately delivered the most tangible value to the workforce.
- Targeted Cherry-Picking: The goal was not to automate 100% of the complex incoming mail, but rather the high-volume, clear-cut standard cases (“low-hanging fruit”).
Description of the Case Study
1. Direct Sampling & Rule-Based Pre-Filtering (Without AI)
Instead of waiting for manual forwarding, the system captures customer emails directly. Unique identifiers (such as the insurance number) are recognized through classic full-text search. This is supplemented by keyword recognition (e.g., equalization allowance, pension account, suspended long-term care benefits). If the customer has exactly one active case in the benefits department, the email is immediately assigned to that case without any human intervention.
2. AI-powered document cleanup
Customers often send emails with countless irrelevant image attachments (e.g., company logos from email signatures). The AI reliably identifies such logos and filters them out so they don’t unnecessarily clutter the document archive.
3. AI-powered detection and cataloging of attachments
The AI analyzes attached documents and automatically extracts relevant metadata:
- DataMatrix reading: The AI recognizes printed 2D codes (DataMatrix) on scanned documents to extract hidden metadata (particularly important for the coded document types used in proof-of-life documents).
- Tagging: Attached documents are analyzed by the AI and automatically assigned the correct document type and appropriate keywords from a predefined catalog.
4. End-to-End Workflow
Identified “standard emails” are automatically converted to PDFs, archived, linked to the customer, and trigger the corresponding follow-up process (task) for the relevant department in the management software — all without any intervention from the KOR department.
Benefits
- Significant time savings: 20 to 30% of all incoming emails are now automatically routed to the correct processes via “background processing.”
- Handling peak workloads: The system effortlessly and automatically processes seasonal surges in documents such as Confirmations of Life.
- Improved data quality: No more overloaded archives due to stored signature images or logos.
- Process acceleration: Departmental teams receive pre-processed, correctly tagged documents and can immediately begin working on the content.
Opportunities
- Scalability: Gradual expansion to include new terms and departments
- Reallocation of resources: Employees can focus on complex issues
- Increased customer satisfaction through faster routing
Risks
- Mismatches caused by incorrectly read insurance numbers (minimized through uniqueness rules)
- Process silos between the archiving system and administrative software (resolved through workflow integration)
In short:
The Pension Insurance Agency has revolutionized its highly manual mail intake process through a smart combination of rule-based pre-screening and AI-powered document processing. By automatically identifying specific customer inquiries and using AI to analyze metadata and document types, 20–30% of emails are now routed to the appropriate departments fully automatically. This significantly reduces the workload on service teams, particularly when dealing with high-volume mail such as life certificates.
Future-proofing your company
