Liberty IDP was set to work in processing emails: Retrieving emails from over 50 functional mailboxes, deriving intent from the content, unpacking the attachments and routing to the relevant department. At the beginning, the solution increased the automation rate from 15% to 30% straight-through processing (STP) – documents that could be completely processed without human intervention.
While this represented an improvement, they anticipated that it was only the tip of the iceberg. A further breakthrough came when a developer from the Liberty IDP team was placed to work in the Baloise Antwerp office for several days each month, to gain a deep understanding of their processes. Now, as well as feeding historical data into the AI model, Liberty IDP was trained differently, using the knowledge of exactly how humans were processing the documents.
Beyond pure AI, Liberty IDP implemented specific rules for certain document types where appropriate, creating a more targeted solution. Processing workflows were split by department (Underwriting, Policy Admin, etc) rather than using one model for all document types.
This approach transformed the effectiveness of Liberty IDP, doubling the STP rate again, taking it from 30% to 65% by 2025. With 11,000 documents being processed each day, this represents substantial cost and time savings. They also gradually reduced their workforce through natural attrition.
The automated system routes documents immediately to the appropriate departments without manual intervention for 65% of all communications. So, 65% of documents no longer have to wait for manual classification and routing – they are now sent to staff in other departments who can get to work on the tasks faster. Eliminating the processing delays has resulted in quicker response times to customer enquiries and claims.
Liberty IDP’s increased automation rates also significantly enhanced processing accuracy. Initially, the system experienced a high false positive rate – documents that appeared to be correctly processed, but actually contained an error. Approximately 20% of incoming documents don’t contain clear reference numbers, so there is still an opportunity to further improve how Liberty IDP finds and extracts these references when they are present, but not easily identifiable.
Through systematic refinement of the AI model and improved training methodologies, the team have managed to dramatically reduce this false positive rate to 2.5%. This is virtually the same error rate typically seen with human processors, so effectively, the automated system is now as reliable as human staff.
This balance of high automation with human-comparable accuracy is a critical success factor. It means Baloise can confidently allow Liberty IDP to handle nearly two-thirds of their document processing without sacrificing quality or risking customer satisfaction.
The ability to maintain such low error rates while processing more than 7,000 documents daily without human intervention (65% of their 11,000 daily volume) demonstrates the maturity and reliability of the Liberty IDP solution in a high-volume enterprise environment.
And, with approximately 300 different document types in use, Baloise are working towards simplifying their classification system which will further improve automation rates.
Baloise employees have benefited from increased flexibility, as staff members have developed a broader range of skills beyond their initial roles. This expanded expertise enables them to take on diverse tasks, extending their contributions beyond document classification. Team members can now manage multiple inboxes, fostering versatility. It also reduces reliance on siloed specialisation within teams.
For customers, the changes have resulted in a noticeably improved journey. Documents are routed more efficiently to the appropriate departments, ensuring quicker resolution of enquiries and claims. Furthermore, communication with Baloise is now processed with consistency, delivering a seamless and reliable customer experience across all touchpoints.