When a customer service task takes too long to complete, there are no winners. From the consumer perspective, “too long,” can be measured by the number of inbox refreshes to receive confirmation that a bank has completed the requested account updates. From the business perspective, “too long,” can be defined as any time spent manually entering or processing information on a form instead of directly interacting with consumers to drive more business. 

To put this in perspective, any bank has dozens of forms available for account holders to open, close, or make changes. These forms range from 1 to 24 pages long, not including any supporting documents like identification or proof-of-residence that may be required for the bank to complete the work. In fact, the largest commercial bank in the United States has 4,964 domestic branches providing these forms for customers to fill in digitally or print out and fill in by hand (https://www.federalreserve.gov/releases/lbr/current/). The amount of time spent manually reading and validating documents quickly adds up, increasing errors and decreasing both customer satisfaction and employee productivity. 

Now consider the time that could be spent up-selling customers on other products or decreasing response and completion times to increase customer satisfaction. IBM’s Datacap front-end document capture workflows along with Automation Document Processing enables businesses to create a powerful end-to-end solution that decreases time and resources spent  manually sifting through documents to refocus attention on customer interactions. 

“How do IBM Datacap and Automation Document Processing help this scenario?” 

First, Datacap makes it possible for all of these documents, regardless if they were emailed, scanned, faxed, etc., to get ingested for use by the bank. If a consumer has a low-quality photo or scan of the form, Datacap can enhance images as well as identify the appropriate form type to determine what account service is being requested (and what business workflow to kick off). 

Once the forms have been ingested, Automation Document Processing can intelligently recognize the most pertinent information on each form and then extract that information for use further down in the business workflow. The out-of-the-box machine learning models recognize common customer and account form fields like a customer’s name, address, and account number. Only fields that are unique to each form type need to be trained for automatic recognition. For each form trained, the system gets smarter and uses that training for subsequent forms. Best of all, training these machine learning models doesn’t require specialized data science skills. This reduction in setup time and hours spent coding or training for each new document type are two significant benefits. 

In addition, Automation Document Processing measures field level confidence scores on extracted information and can request human intervention in error situations. This helps businesses maximize automation and minimize risks. 

“That’s great, but I’m not a bank.” 

The same scenario above could easily be applied to an insurance company that processes account quotes & applications or claim forms. In addition, federal or state government departments can apply Datacap and Automation Document Processing to accelerate enrollment and servicing of social services programs or state pension / retirement plans. Using Automation Document Processing, these services programs can identify key fields on applications and correlate with new information or existing submissions to determine eligibility for program benefits.  

With the explosion of content and remote work, not only is valuable information stuck inside documents but maintaining customer satisfaction is increasingly challenging. Stay ahead with the Cloud Pak for Automation, IBM’s comprehensive Automation Platform. Clients can leverage their existing Datacap investments for document ingestion, transformation, and separation and apply Automation Document Processing for AI-powered content classification, data extraction and enrichment, and validation.