OCR … OC WHAT? What is Optical Character Recognition?

Optical Character Recognition or OCR is a technology enabling computers to “read” text off of documents. It is usually priced per page and can be riddled with mistakes. The lack of accuracy makes it of little practical value for organizations looking for a clear picture of the data in the documents. In order to get the true picture, there’s often a team of people, also known as “human in the loop (HITL)” that need to augment the accuracy and completeness of the result (and those humans need to be paid, so costs go up).

OCR has a long way to go, especially in the auto industry where forms are constantly changing and new document types are added regularly. However, by combining state-of-the-art OCR technology, with applied intelligence ”learned” from a vast contributory database and subject matter experts, solutions like InformedIQ can intelligently “read” the document, turn it into actionable insights and perform complex calculations.

It is important to accurately parse the data. With the power of AI, you can get accuracy rates of 99% on known document types. For lesser known document types, accuracy rises quickly over time. Until the “machine’s” accuracy is where you need it to be, you control, based on your lending policies, which documents you feel comfortable accepting and others you would like to manually review.

Proof of Income documents are so widely used that AI models can extract data, classify the document, compare the data and determine if the document is based on a fraud template found on the dark web in addition to calculating the income based on your credit policies.

Quickly processing a consumer or auto loan is the ultimate goal of most lenders, so the more documents that can be instantly verified, without human input, the sooner the lender can make their decision. Lenders are able to pass a loan that historically would have needed a person to manage it through data entry, classification, verification, calculation, compliance, or a fraud alert. For the small number of documents that actually do require human manual review, the system should let the reviewer know exactly what pieces of information they need to review – ie is a signature or required document missing?

Image quality is another complaint about OCR providers. If the image quality is bad the accuracy will be low, and lenders do not want to invest in new processes or infrastructure to manage this since more and more documents are becoming digitized. Offering better ways to instantly receive digitized documents improves the consumer and dealer experience and addresses image quality issues.

OCR also cannot determine loan defects. OCR capabilities with the addition of machine learning ensures contracts are complete and accurate. The ability to collect documents ensures all the lender’s needs are addressed without disrupting their current workflow and providing the best solution from one financial technology provider. More options lead to customized streamlined solutions with better business outcomes.

Transforming documents and data into actionable insights and decisions is no easy task but we simplify it for you. Our machine learning models have been trained by millions of documents to support the auto, consumer lending, and mortgage loan industries.

Learn more on how to reduce front-of-house labor costs while improving accuracy and reducing errors. Request a Demo.

As Featured in American Banker