If you are a buyer in a company looking for automation software, you will come across “OCR” eventually. Analysts recommend looking for “OCR software”. And if you read up on it, “OCR software” will do all the magic heavy-lifting related to documents. It can classify documents, extract data from them, and understand them. And of course, it is all artificial Intelligence and you don’t have to configure anything. Pure magic!
What they sell you as “OCR software” is usually a Capture platform, a BPM or RPA platform that uses, among many other technologies, OCR.
OCR stands for Optical Character Recognition, or as the more experienced people often say “Occasionally Correct Recognition”. OCR does nothing more (and nothing less) than looking at all the pixels of an image file and turning them into characters and words. That’s really it. That step alone does nothing much for your automation. Your goal is not to scan a document and then have text that you can copy and paste somewhere else. Your goal is for it to understand what the text means, where the piece of data is in it, that you are looking for, or to understand what is an invoice and what is a credit note, based on that text.
Calling the entire automation platform an “OCR product” doesn’t sufficiently describe it.
OCR is absolutely important when you digitize your paper documents in order to automate your processes. In fact, without OCR you wouldn’t get very far. But it is not the most exciting part by any means. These days, OCR has become a commodity. There are dozens of OCR engines for integration in software, as well as cloud services that return text when you send them images. But that alone doesn’t understand a document or create actionable insight.
After using OCR to turn an image into text, most modern document automation products apply a series of additional steps, like classification, data extraction, business rules, human review, formatting, aggregation, conversion, and export. These additional steps do the actual understanding and trigger the automation you are looking for. These parts, especially classification and extraction, are where the various vendors can shine. Classification and data extraction make use of Machine Learning and other Artificial Intelligence techniques to do their task with minimal configuration by you.
So, calling a Capture or Document Automation product “OCR software” really is like calling the latest Tesla model a “wheel thing”. Yes, a Tesla has wheels, but they aren’t what makes Tesla shine. Every car needs wheels. We expect that of a car. And so do we of an Intelligent Document Automation platform.