New study: The Agentic AI gap
Read now
LKUF: Around 90% of billing documents processed with AI support
Challenge
Every year, the LKUF processes thousands of reimbursement claims from insured school teachers in Upper Austria. Unlike the standard health insurance model, insured members pay their medical expenses upfront and submit invoices along with other documents such as prescriptions, applications, and claims for reimbursement. Until recently, all these documents were reviewed and entered into the system manually by the claims staff.
To process incoming documents correctly, insured members needed to pre-categorize their submissions, for example as a “dental invoice.” Especially with more complex documents, the correct classification is not always obvious. Missing or incorrect data leads to frequent corrections and additional effort in the billing process.
Solution
LKUF became aware of Cloudflight through a reference from an earlier project. Together, we kicked off a proof-of-concept in mid-2025. In a focused workshop, we helped structure the client’s many ideas and create a realistic picture of what AI can deliver.
- The result is a targeted data extraction system: Building on a pre-trained open-source GPT model, the team applied targeted prompt engineering, tailored to the specific document structures and supported by structured pre- and post-processing logic. This ensures reliable and consistent data extraction.
- The model runs on LKUF’s own server infrastructure. The GPU computing power required is sourced securely and flexibly from an EU data center.
- Both scanned postal submissions and digital online submissions are processed. Extracted data flows directly into the existing database.
- The collaboration between LKUF and Cloudflight is built on agile principles. Continuous alignment ensures that adjustments and improvements find their way into the solution quickly.
Benefits
The deliberately lean approach has paid off. Rather than packing in as many features as possible, the team focused on the core pain point. This allowed LKUF to move from initial concept to live operation in less than one year. The system went into production with postal cases first, followed shortly after by digital online submissions.
- Up to 90% of cases are now prepared with AI support, significantly relieving claims staff and minimizing manual preprocessing
- The solution extracts more data points than before and reliably processes paper-based and unstructured documents alike.
- Correct digital submissions through the myLKUF portal were already flowing automatically into the billing system, but the new AI solution takes things a clear step further.
- Experience from similar projects formed the design: clear system boundaries were defined from the start, ensuring the AI tool fits cleanly into the existing process chain without creating conflicts of responsibility.
Our partner
LKUF is the health and accident welfare fund for compulsory school teachers in Upper Austria, based in Linz. In addition to standard direct billing, LKUF operates on a cost reimbursement model in the outpatient sector, serving thousands of insured educators across the region.
The organization combines its public service mandate with a genuine drive for improvement, actively pursuing technology solutions to advance internal processes and the service experience for its members.
Other successful projects
voestalpine myAPP reaches 12,000 employees with central information and strengthens cohesion
For voestalpine, Cloudflight created myAPP – a tailored app that connects 12,000 employees worldwide. It delivers central information, safety alerts, and interactive features that make daily work easier and strengthen team cohesion.
