Case Study: OCR & NLP
A.I and OCR platform, DocIntelligence helps leading Sydney Fintech Company save over $100,000 a month
Background
Major data extraction bottleneck hindering growth.
A leading Sydney Fintech company had identified that they had a major bottleneck in their application process that was hindering the number of new users they could sign on. Consequently, it was slowing their growth as they were struggling to keep pace with the amount of monthly signups.
​
Under Australian regulation, the company had to extract a large number of key fields from a range of confidential documents including payslips, mortgage statements, licences etc. They had tried to utilise standard Optical Character Recognition (OCR) to extract the fields required, but it had not been effective on documents with a high degree of variation - such as the mortgage statements or payslips.
As it stood, each application was taking up to 2 hours to process.
​
The company approached Remi AI to see if our latest OCR and Natural Language Processing (NLP) platform could help them extract information from the desired fields on relevant documents, thus reducing application time.
​
​
Remi AI connected the latest version of the OCR platform, DocIntelligence, into the clients Application pipeline and immediately set up the custom A.I modules to begin training on the most complex documents.
​
Within 4 days, the A.I modules in DocIntelligence - OCR, Neural Networks, and NLP - had all been trained to between 86%-95% accuracy. This was 6% higher than the APIs the client had tested previously.
​
These were then brought into production in the applications pipeline.
Approach
A 2 hour process reduced to 2 minutes.
Results
With accuracies as high as 95%, DocIntelligence reduced the time spent extracting data from the applications from an average of 2 hours to under 2 minutes. This allowed the Applications Team to increase the number of applications they were processing by more than 6000%.
