When it comes to all the cool innovations that are driving simpler, more flexible and more effective automation for routine, high-volume business processes, there’s one that seems to get lost in the shuffle as artificial intelligence (AI), robotic process automation (RPA) and machine learning compete for the spotlight. The stepchild to all of these buzz-worthy technologies is one that stands to deliver the greatest impact by leveraging the other three—because it attacks the most painful part of any business process: entering data into business systems that may come from a document, email or other unstructured source.
That technology isn’t brand new, but it has greater relevance and impact today because it has grown more accurate, more flexible and easier to integrate into an existing business technology ecosystems.
We’re talking about intelligent data extraction aka intelligent document extraction. Regardless of which term you prefer, IDE focuses on finding relevant, actionable information hidden within unstructured content. That means finding the invoice number, vendor name, PO number or purchase amount within a scanned paper or electronic invoice. It may mean parsing the line item details of a sales order and extracting all of the inventory item details to input into an ERP for forecasting, inventory or buyer analysis.
Often, intelligent data extraction is linked to OCR or optical character recognition. By now, OCR is a mature, well known and not very sexy technology. It dates back to the days when futurist Ray Kurzweil was focused more on helping the seeing impaired to read than on helping Google with developing natural language processing technology for Google.
In that sense, intelligent data extraction comes from a long line of technology innovations intended to solve some fundamental business problems.
And today, it is more relevant and impactful than ever. Here’s why.
Before we get into why other buzz technologies make intelligent data extraction so valuable, we should pause for a moment to issue a ‘buyer beware” alert. That’s because, like some other forms of supposedly intelligent automation marketed today, some solutions are really relying on humans working behind the scenes to do the work other humans prefer not to do.
Some outsourced solution providers extract data from your documents behind the scenes manually, either executing the data extraction entry manually, or for third-party data verification. In this scenario, there may be a significant time delay in getting the data you want, because the extraction is really performed the old fashioned way, which may take between 24 to 72 business hours. Really, this represents a company masquerading as an intelligent process provider—and that’s NOT what we’re talking about here.
In the scenario we’re describing, where technology does the work faster, cheaper and smarter, intelligent data extraction technology relies on OCR to transform data from scanned images that may be uploaded to the system, or via native digital files that might be received via email. These systems interpret information source, identify and extract the appropriate data. Next, it can integrate with an existing ERP, ECM or other business system and enter the data in an appropriate structured data field.
But, you might ask, what if the system can’t read a character because the source file is corrupt? For example, a coffee spill or stain on a scanned paper invoice.
This is where machine learning comes in. Machine learning often involves having a system identify an error or exception that it can’t handle. To do so, it has to notify a human process owner, who may have to intervene to identify a character or word, or to cross reference another business system to validate a data field. But, rather than relying on a human to do this every time, the system can monitor and record the human operator’s actions, and update its own algorithm. Without any coding or configuration.
Here’s what that means in real world terms:
For companies stuck in the status quo, intelligent data extraction, combined with machine learning and low code/no code solutions will make it increasingly difficult for companies to do things the old fashioned way and much easier to say yes to innovation in a way that is scalable and budget-friendly.
To learn more, visit the Artsyl Resource Page at