Top Pet Insurer Achieves High Accuracy with Poor Quality Documents





Process Automated



Document Classification Accuracy


Data Extraction Accuracy Including Poor-Quality Documents


Page Processing Time

Client Profile

The client is one of Europe’s largest providers of pet insurance. The company realised the importance of the information found in most of their business critical documents like medical claims. And, like many companies that run on huge volumes of documentation, it was struggling to extract all the data in all the documents it received.


The insurer receives invoices, veterinary reports, pre-authorisation and claims forms from thousands of suppliers, clinics and individual customers, often in non-standard formats or formats unique to each business. The readability is frequently poor.

Most automated document processing systems were unable to processing this unstructured data, but extracting it manually was a time-consuming, expensive and error-prone process where useful but non-essential data wasn’t digitised.

So the insurer looked to process more claims without increasing staff while accessing the highly valuable insights locked in that unprocessed data to build up a picture of its pet owning customers, respond to them more quickly and offer better services.


AntWorks created a synchronous data extraction system for the client linked to instant messaging whereby when a policy holder or a veterinary clinic submits the claim online, AntWorks captures the claims data and seeks confirmation and approval in real time.

AntWorks also helped develop a new process that sorts and classifies medical records, pre-authorisation forms, invoices and claim forms, automatically sending the invoice data to the Insurer via an API for downstream processing.

Processing accuracy at the proof-of-concept stage despite a high proportion of hard-to-read documents. As the solution goes into production and processes more material accuracy levels will rise further.


“The client was very impressed by the high level of accuracy at this early stage with small training sample sets, especially given that the document quality was often poor,” says AntWorks’ sales manager, Natasha Grinstrit.

“They were also pleasantly surprised by AntWorks’ commitment to the success of the project and relied on our expert knowledge. From our point of view it’s simply a given that a project of this level of complexity requires us to collaborate very closely with a client.”

The client was more than satisfied with the proof of concept and has moved forward to production.