Using AI for invoices lets ControlExpert add structure to data
By using Abbyy software and OCR technology, ControlExpert can automatically process and classify millions of structured and unstructured documents and invoices.
ControlExpert, based in Germany, is one of the largest international automotive insurance claims processors in the world.
Each year, customers send millions of documents -- some handwritten, some neatly typed out -- to the company. To handle such vast amounts of data, ControlExpert has automated much of its workflow, including the reading and filing of those documents. The company uses an amalgamation of in-house and third-party tools to do so, but recently, more and more, it has turned to AI for invoices and Abbyy optical character recognition (OCR) technology.
"We receive documentation in all facets," Harlan Vold, vice president of technology and operations at ControlExpert, said. "Most of it's unstructured data, so it's really critical that we can straight-through process it, straight into an actionable state," he said.
Abbyy AI tech and products
Abbyy, he continued, has been instrumental in making that happen.
Founded in 1989, Abbyy, a software vendor based in Milpitas, has developed a diverse portfolio of products and technologies focused largely on language. Products include digital dictionaries, translation products, OCR platforms and robotic process automation tools.
The OCR platform FlexiCapture has been particularly useful for ControlExpert, which processes many different types of documents.
[For our unstructured data] it's really critical that we can straight-through process it.
Harlan VoldControlExpert
It's a powerful platform. Using OCR technology and AI for invoices, FlexiCapture can automatically capture and extract data and text from structured or unstructured documents. Abbyy AI and machine learning technology enables the platform to automatically classify extracted data, as the platform automatically learns from user interactions how to effectively use the data.
Abbyy tools also work with digitally born documents -- automatically extracting and sorting data found in digital documents or email and making that data easily searchable. It's an important feature in an increasingly paperless world.
Getting easier
Using and teaching the AI and machine learning components in FlexiCapture -- and in other Abbyy AI products -- is relatively simple to do and will get simpler in the coming months, Anthony Macciola, CIO at Abbyy, asserted.
Currently, Abbyy is focusing mainly on simplifying user interactions with its products, Macciola said.
Automating the process of structuring unstructured data from an invoice, for example, traditionally, has "required a relatively high level of sophistication from the user," he said.
"I think, in 2019, we will simplify that to the point that anyone, after they see a two-minute video or vignette, will be able to do that themselves," Macciola said. "So, I think the adoption of content intelligence will become much more pervasive."
In 2018, Abbyy released its latest iteration of FlexiCapture -- FlexiCapture 12 -- which added more, and the company said easier to use, AI for invoices features, automation and cloud computing capabilities.
According to Vold, these latest steps have enabled ControlExpert to move away from developing its own technology, which it had supplemented with Abbyy AI products, and "move towards going to Abbyy" entirely.
Vold noted that, for FlexiCapture, some of the handwritten invoices "are not necessarily capture-able at the moment."
That's a challenge for ControlExpert, as some invoices "basically say part number, an amount plus tax and a handwritten signature." Future product iterations are expected to solve this issue.
