Challenge
Daily LCM receives approximately 100,000 certificates from doctors, dentists and physiotherapists, which require an intervention in the costs incurred by the patient. An important part of the certificates is handwritten, and the variation in printed certificates is such that the classic optical character recognition (OCR) solution could only recognize 35% of the “nomenclature” codes and 11% of the treatment dates automatically. The other fields, as well as the unrecognized nomenclature codes and treatment dates, had to be read manually and retyped in an application. This involved several hundred people and also meant a loss of time of an average of five days to make a refund.
The objective of the project was to achieve a higher recognition rate with artificial intelligence (AI) in order to reduce time loss. After all, faster repayment means great satisfaction among members. At the same time, implementation costs and operational costs can be reduced.
Solution
- We built a bot that automatically reads doctor’s handwriting on sick forms and created an AI algorithm that surpasses the human ability to read bad handwriting.
“We went from a difficult and time consuming process to an efficient, AI based system which even deciphers bad doctor’s handwriting for us!”
Civil workers gain great advantage of the AI that deciphers the doctor’s handwriting and suggest what’s written before it automatically puts it in the system. They can now focus on other core tasks and above all, the error rate of unknown causes is drastically reduced. - Speeding up the process of registration
- Improved the quality of the electronic sick forms by reducing the unknows
- Lower cost of backlog management and more focus on core jobs