CASE STUDY
Delivering millions in savings for a mortgage title & deed data provider
Cognizant is a global IT systems integrator and consulting company. Cognizant is an industry leader in Intelligent Automation and was seeking a solution that could meet the stringent accuracy requirements while decreasing the resources required and shortening the turnaround time for processing.
In the context of unstructured data, the ability to give feedback to the model, the learnability part of the tool, is one of the most important things, otherwise; we are back to the old template-based approach because you have to go back and retrain the model.
-Abhinav Kolhe, Senior Director, Digital Strategy and Automation Solutions with Cognizant
40M
documents processed yearly
40%
reduction in processing costs
Problem
Cognizant’s client, a prominent US real estate data provider, provides comprehensive data about real estate transactions across the US and licenses this unique data asset to critical residential real estate market players. Competition in this segment is fierce, as these documents are all publicly available. Key differentiators are the speed at which the data becomes available in a structured format and the accuracy. Historically, this problem has been brute force manual data entry, and Cognizant deployed thousands of offshore resources to process this vast river of documents. As a strategic partner to Indico, the challenge is automating identifying and extracting relevant data from unstructured mortgage title and deed documents published across the US by local municipalities – up to 40 million of them each year.
Solution
Cognizant selected Indico, with their Intelligent Process Automation (IPA) Platform helping the client achieve 40% annual savings in processing costs. Indico’s IPA Platform is based on its proprietary Transfer Learning approach, allowing non-technical users to quickly build custom machine learning models tailored precisely to the customer’s needs. Using a point and click interface, SME’s can upload their training samples, create custom attributes for extraction, classification, etc., and quickly build models with Indico’s proprietary AI-assisted labeling. Most importantly, the solution had to have enterprise-scale capabilities. Indico’s cloud-native microservices architecture enables continuous optimization of the compute footprint and can scale up to any throughput or processing volumes.
Results
A key differentiator of the Indico Platform is that the models gain accuracy over time. Cognizant’s goal was to augment the employees who are experienced in the titles and deeds process, not completely replace them. Indico enables these process experts to review results, with a pane that shows the level of confidence for each field as to whether the model is producing accurate results. For fields with, say, 70% confidence level, process experts can help teach the model how to increase accuracy. Eventually, those fields with high confidence ratings may be hidden from the review pane, thus speeding the process further while boosting accuracy ratings.
Share on social:
Get started with Indico
Talk with us
Discuss how the Indico Platform can help you tackle your unstructured data problems