Businesses spend billions each year capturing transactional information and metadata. The e-commerce wave promises to reduce some of these costs by turning transactions into automated electronic processes. Yet most businesses are still exchanging paper documents. As a result, they still spend billions on data entry, filing, searching for documents, copying, faxing and other forms of paper shuffling.

In the effort to automate paper-bound transactions, advanced document imaging systems represent a kind of precursor to e-commerce. Once a document is captured as an electronic image, it becomes a searchable, shareable resource that can be accessed from anywhere via the Web. Images can also be mined for their data in order to drive business transactions.

While early imaging applications relied on key-from-image techniques that were only a step or two ahead of paper-based data entry, new and proven technologies are helping cut unnecessary labor from document imaging processes.

A Midwest Utility Automates Invoices

Character recognition technologies such as OCR and ICR have long been used to reduce manual data entry costs in applications involving structured forms. Loan applications, tax forms and surveys are a few examples of document types that have been successfully automated with minimal OCR/ICR validation required. These forms have consistent layouts, enabling users to define fields in which particular types of information can be found. Traditional forms processing technology works best, for example, if the “total amount” field consistently appears in the bottom right hand corner of a document.

Consistent forms, however, represent only a part of the document volumes that many organizations have to process. Invoices, for example, are often the lifeblood of transactions, yet they vary from vendor to vendor, so they can’t be automated with conventional data capture software. Over the past three years, a great deal of progress has been made in automating semi-structured forms (or “variable forms,” as some call them). [Editor’s note: the term “unstructured documents” is usually applied to less form-like documents such as correspondence and resumes.]
Early adopters of variable forms processing technology typically cut their data entry costs by at least half. American Electric Power (AEP) of Columbus, OH, is no exception. The utility recently installed a variable forms processing system, and early tests point to a return on investment within 13 months.

AEP is no stranger to imaging. Over the past seven years, the company has been capturing some 2,500 documents per day. The majority of those documents are invoices that arrive from nearly 200,000 different vendors.

“We receive invoices for everything from cotton gloves to the equipment used in our power plants,” says Ken Jones, senior support specialist for AEP. “The amounts can be from a few pennies all the way up to millions of dollars.”

AEP’s images are stored in a FileNet repository, and until recently the capture step was handled by FileNet’s Panagon frontend. Because the invoices have variable layouts, AEP has always used key-from-image data entry to capture the vital data from each image.

“We were originally introduced to OCR technology at a FileNet users’ conference,” says Joe Boyden, senior IT architect for AEP. “However, the application we saw there was template-based [meaning it was designed for consistent forms], and we knew that would never work for us.”

Boyden says he first learned about free-form technology in the pages of Transform Magazine. Specifically, he noted a 2001 award for AnyDoc®INVOICE from Tampa-based AnyDoc Software (formerly Microsystems Technology).

“I saw that [AnyDocINVOICE] had been awarded “Transform Product of the Year” for [variable] forms processing and realized it might be the answer we were looking for,” Boyden says. “We spent a year discussing the idea before signing the contract. After about a month of development work, we began a phased implementation, which we spread out over another month.”

AEP is now using the former AnyForm for Invoices product, now called AnyDocINVOICE, to capture up to 11 fields on each invoice, including information such as invoice number, date, vendor and gross amount.

“AnyDocINVOICE works best on invoices that include a purchase order number,” says Jones. “Once the software identifies the PO number, it can compare the data on the rest of the invoice to the purchase order data already in our system.”

Jones estimates that about half the invoices AEP receives include a PO number. “A very high percentage of those invoices could probably flow through unchecked because the accuracy rate is so high,” he says. “However, the people using the system do not have the confidence yet to let that happen. They are still running quality assurance checks on 100 percent of the scanned invoices.”

Despite this stringent process, AEP has already been able to cut its data entry staff in half since installing AnyDocINVOICE. “We expect the installation to pay for itself within 13 months,” says Boyden. “We hope to be able to further reduce our data entry staff, but first we need to raise our users’ confidence level to meet the capabilities of the product.”