Content Management and Capture

 View Only

How anyone can use deep learning to turn unstructured content into actionable data with IBM Automation Document Processing

By KATHRYN TIRADOR posted Tue October 13, 2020 09:22 AM

We understand automation will make our businesses more profitable and free up time to focus on high value tasks. We also know that data scientists can build sophisticated models to make this happen, but they are in high demand and have too many projects to tackle. Meanwhile, businesses continue to operate and are ingesting millions of documents with data locked inside of them. Who has the time or resources to manually read all of these documents and enter data into a system, or spend 6 months to a year setting up a traditional capture system for a new process?


IBM Automation Document Processing empowers business users and knowledge workers to quickly deliver their own processing solutions with accuracy thanks to AI with deep learning.  IBM Automation Document Processing can read, refine, and apply high-quality data to workflows and applications eliminating manual document processing. But you may be wondering...

You said AI with deep learning? Doesn't that require a data scientist? Not with IBM Automation Document Processing. We've done the hard work and created pre-trained models for popular document types like purchase orders, utility bills, invoices, and more. If you choose, you can easily re-train the model with your specific documents with a no-code user experience. It provides a step-by-step process to guide and instruct  you on how to choose the best model with the highest accuracy.  With few samples, you'll have a model that automatically classifies and extracts key data from unstructured documents, the first step towards eliminating manual document processing.

How do I deliver my own processing solution? Doesn't that require an app developer?  Low-code tools make it possible to build and deploy a document processing application without a ton of help and resources from IT or a specialized app developer. As part of the IBM Cloud Pak for Automation, IBM Automation Document Processing provides out of the box templates that enable users to build an application with a drag and drop design experience. 

Ok, I get that it's easy to set up, but what about the actual processing of documents?  Your new document processing application will automatically classify and extract the data that is most important to your business and flag any fields where it identifies a lower confidence level. Users can verify classification and extraction  results that have been flagged and easily correct them in real-time, ensuring that only the highest-quality data is used to power mission critical workflows, downstream applications, or drive your RPA bots.  Since this application was easy to set up, you can start automating previously unaddressable processes The deep learning models use a concept called 'transfer learning' to use existing knowledge to quickly understand new document types. This enables you to continuously expand your automated document processing operations with the same no-code user experience. With this deep learning paradigm, the possibilities are endless!

Where are the documents and data kept after processing?  The solution is fully integrated with FileNet Content Manager, simplifying document and data storage by applying your existing filing architecture and business rules to each processed document. The content and metadata are automatically saved in FileNet within the appropriate document class. FileNet and IBM Enterprise Records provide robust governance and records management capabilities to make sure your business isn't exposed to unnecessary risk and can comply with legal and regulatory requirements. If you already have large investments in another content services repository, the good news is you can also use your existing repository to store documents, while still submitting documents for processing from anywhere. 

Need to see it to believe it? Check out the Automation Document Processing Solution Overview video.

We are excited to announce this new capability today that will be generally available in December 2020. Don't just take my word for it though, see what our customers and partners have to say.

"IBM Automation Document Processing provides the ability for setting up document classification and data extraction powered by AI and deep learning which will significantly reduce the need for manual human input and intervention." - Jack Moore, Director, Solutions and Product Management, DHL

"The guided workflow will provide a smooth user experience and key contextual information about the progress of the project. With the integrated data validation workflow, we believe the extraction accuracy will increase significantly." - Scott Power, President & CEO, Magiclamp Software 

"We believe that these AI-based classification and extraction tools are the future of the space and that IBM is an industry leader." - Eric Walk, Senior Technical Architect, Data Solutions, Perficient