Overview

Although Web Services and Electronic Collaboration are the key buzzwords of today’s economy, over 70% of business transactions are still conducted via paper and fax causing many organizations to rely on intensive, manual processing.  For example, supplier invoices and customer orders are still received via fax machines of yesterday.  The process for handling these documents is time-consuming, error prone, and most of all unnecessary.  Most organizations handle a large volume of paper based documents especially when dealing with Orders and Invoices.  The process today is one that has been in existence for decades.  Once a fax or letter is received, a group of employees manually enters the information in the corresponding back-end application such as SAP or Oracle. 

Although this is the status quo, many companies still deal with:

• High costs associated with manual data entry
• Lost orders and invoices
• Errors from manual key strokes
• Lengthy processing times, inhibiting customer service
• Time consuming reconciliation of transaction changes
• In ability to automatically acknowledge and verify receipt of documents
• High turnover rates of data entry clerks

Changing the way in which trading partners interact with you is an evolving and slow process, but changing the way your organization deals with paper transactions can be done in a few weeks. BPM combined with Artificial Intelligence is removing the manual and cluttered processes of yesterday for many organizations around the globe.  New standards such as BPEL (Business Process Execution Language) are making the transition possible.  Today, BPEL based designers allow organizations to run and monitor long running processes which may take days to complete.  More importantly, these new BPEL engines can easily incorporate “Human Interaction” which is critical in automating the human “user exits” which exist in virtually all processes.  

Manual Data Entry – The Status Quo

The Input Management Problem — Enormous quantities of printed and electronic information flow into businesses, government agencies and other organizations on a daily basis. Most of this information comes from customers, suppliers, employees and other third parties conducting transactions or corresponding with the organization through the use of paper and electronic forms and documents such as letters, resumes, new account enrollments, credit applications, tax returns, invoices, legal briefs, and regulatory filings.

Pertinent information from these forms and documents must be extracted, perfected, formatted and exported to an organization’s information management systems as cost effectively, accurately and quickly as possible. Processing delays, incorrect or invalid data and inefficient methods can adversely impact an organization’s revenues, operating expenses and interactions with customers, suppliers and employees.

Current Approaches are Inadequate — Traditional methods for capturing data from paper forms and documents rely principally upon manual data entry. According to current analyst reports, organizations in the United States spend $15 billion a year manually keying data from forms alone. Organizations generally perform this work internally or outsource it to BPO (Business Process Outsourcing) firms. But both approaches have inherent problems. Organizations performing this work internally usually distribute it on a departmental basis, where highly paid knowledge workers manually key data directly into information management systems with minimal data validation and low productivity rates. The accuracy of the information and processing times usually suffer, and costs incurred from both erroneous data and poorly utilized human resources are high.

Analyst studies:

o        25% of enterprise paper documents will be lost and never found (Gartner)

o        95% of enterprise information is still on paper (IDC)

o        Enterprise Content Management market expected to be 3.8 Billion in 2007 (IDC)

o        80% is the percentage of documents in an enterprise that are unstructured data, and it is growing at 200% a year (Yankee Group)

o        $8 is the average cost to process an invoice. As much as 70% of all invoice processing costs are document handling and data entry process (International Accounts Payables Professionals)

o        86%, is the percentage of invoices still printed on paper (Gartner)

Moving Into the 21^st Century

While we can’t always get away from paper transactions entering our four walls, especially with our smaller trading partners, solutions exist today that can have a dramatic effect on how we handle these documents inside our four walls.  These solutions must not just provide for simple scanning and archiving.  Today real Return on Investment will be achieved with solutions that include the following capabilities:

• Scanning paper forms and documents to create digitized images;
• Performing image quality assurance functions;
• Optimizing images for the application of recognition technologies and viewing purposes; 
• Applying optical character recognition (OCR), intelligent character recognition (ICR), barcode, and other recognition technologies to extract data;
• Data Validation via FreeFormation technology which removes the need for hundreds of image templates.  This is a key factor and includes the use of (AI) Artificial Intelligence which can automatically learn where data is located.  The only true way to reduce manual data entry is with AI, not templates.  Although documents such as Purchase Orders and Invoices generally contain similar information – their format and layout can be drastically different;
• An event and data driven workflow for routing and tracking all data and images; and
• Back-End Application Integration to both Document Management Systems and Enterprise Resource Planning system.

It is interesting that when you analyze the solutions in the marketplace today that most of the questions from end users relate to the accuracy rate of the OCR engine.  Well it is true this is a major requirement – the OCR engine can only bring so much efficiency.  The real benefits of a solution lie within a solutions Data Validation & Workflow capabilities.  These are the components that truly save time, both on the front end set up and back end operation.

Data Validation – Free Form Recognition to the Rescue

While many solutions in the marketplace claim to have data recognition and validation, it is important to look under the covers of these solutions.  As stated previously, optical character recognition is only a portion of the necessary technology.  Over the past few years, a few organizations have released what is now called “Free Formation” which applies artificial intelligence to the recognition of data.  Remember, even though an Order has common data elements they will typically differ in both layout and field design.  For example, the Purchase Order Number may be located in the top right hand corner on one document and in the left hand column below Customer Name in another.  If you had to design recognition templates for each customer, your design team could take months to accomplish the learning procedures necessary for hundreds of document versions.

Free Formation removes this template process by focusing complex algorithms and automated learning procedures that incorporate the following principles of interpretation and validation:

• Key Words - Identify: “invoice number” or “inv. no.” or another phrase
• Learning Repository - Cross-check: for example, partner master data, historical purchase order data, results, or previous experience from earlier processing of a document
• Plausibility & Patterns - For example, if the invoice number is identified, then it is probable that the date next to this number is the invoice date

BPM: Workflow Removes Process Inefficiency

Process automation is the second critical factor when rationalizing the move toward the removal of manual processes. So how can BPM and workflow help your organization?  Let’s take a look at a specific business process: Procure to Pay.  While ERP has led to advances in the managing of overall transactional data – there are still issues that are caused by errors outside the firewall.  For Invoices, this can include:

• Missing or Incorrect Information in the document sent by a partner
• Duplicate Invoices
• Price Discrepancies (Invoice doesn’t match the system generated PO)
• Quantity Discrepancies (Quantity of goods doesn’t match system generated PO)

If it takes you 5 minutes to transition an invoice from paper into an electronic format, it is probably taking you days to fix the mistakes that exist in upwards of 30% of all inbound documents.  Why?  Routing a document and collecting the appropriate approvals is a cluttered and expansive process that requires multiple people to route, cross-check and approve discrepancies. 

All to often we under estimate the complexity of what we would consider standard and simple transactions and forget that for each document such as an invoice there are up to 15 or more internal checks that must happen in a logical order for that invoice to be posted and paid.  A standard invoice management process would include the following sample stages, of course the process could take up pages based on complexity:

1. Invoice received (if electronic sent on, if paper sent for manual data entry)
2. Three-Way Match if ERP system is capable
   1. Have goods arrived (Quantity check)
   2. Does Invoice Amount match PO
3. If Match post to the ERP, if Exception proceed
4. Internal Exception Management
   1. Decipher error
   2. Manually decide which department

                                                              i.      Based on amount send to appropriate manager

3. Send email and route message to log into the system
4. Recognize exception and resolution
4. External Exception Management
   1. Send back to original supplier with notice of error
   2. Call supplier and have cross check their system
   3. Supplier Confirm or Dispute
   4. Resolution
5. If Confirmed, Supplier re-issues corrected invoice
6. Posting to Back-End Application for payment

It should not be shocking that there are multiple bottlenecks in this process if it is not automated and tracked with a comprehensive workflow solution.  These delays cause organizations internal havoc: buyers can not see outstanding balances as they are not posted in the ERP system, accounts payable may not be able to receive payment discounts (n/30), and supplier relationships suffer due to insufficient communication and resolution.

What BPM tool should you consider?

When it comes to workflow, there are a lot of options for an end user to choose.  Some organizations have used the programmatic webflow languages such as that provided in older SAP systems and others have purchased workflow solutions from integration vendors that allow for designation of certain actions such as document routing.  Although these solutions have offered some basic business value, the architecture has had limitations including:

·         Hard Code Programming – older solutions are not designed via drag and drop functionality but rely on the generation of code

·         Programming Lock-In – many workflow solutions have been built on proprietary code sets

·         Complex Workflows – coding multiple request/response operations with lack of correlation to visualize a multi-step process

A new open standard is helping to resolve many of the issues that older and proprietary solutions could not handle.  This standard known as BPEL (Business Process Execution Language) has been selected by key vendors including IBM, SAP and Oracle.  While it is not appropriate to go into all of the architectural details, BPEL is an ideal language for coordinating long-running interactions consisting of synchronous and asynchronous message exchanges. 

If you look at the multi-stage process for posting invoices into an ERP system, it is logical to classify this entire series of events as a long-running process where multiple requests for information may be required from both human interaction and application table and value look-ups.

BPEL is specifically designed to handle both underlying architectural requirements of the potential asynchronous and synchronous exchanges that may need to take place.  For example, if the user is at a browser and is waiting for a response after putting in a data request, a synchronous call may be fine.  The rationale might be that maintaining a HTTP connection is reasonable for the short duration of this required action (i.e compare PO amount to Invoice amount by doing a search of ERP data).  However, a different usage scenario such as management approval of a discrepancy might have user feedback happen via e-mail with a response time measured in days instead of seconds. 

When business processes require multiple actions among requests that may not have a response for days to weeks, architects must turn to tools and standards that allow them to incorporate multiple methodologies in an easily definable and executable language.  BPEL has unique advantages including:

• Support long-running processes and the need for reliable, asynchronous communication
• Support for Web Services and Service Oriented Architectures
• Endpoint Management of a service
• Process Persistence
• Transaction & Messaging Correlation

Although any of the above points can be achieved by writing code in a general purpose programming language, the primary reason for turning to BPEL is that it provides abstractions and infrastructure that allow analysts to model and architects to implement solutions more efficiently.  In the end, the solution should not be more complex than the original problem.

Business Take Away

• Upward of 70% of all business documents including Orders and Invoices are still communicated via manual processes including faxes and direct mail
• Organizations can improve internal process efficiencies by utilizing artificial intelligence to reduce manual data entry.
•  Newer workflow standards allow organizations to automatically route information for approvals and cross-checks, eliminating long delays and potential errors.

Technical Take Away

• Solutions must go beyond general OCR (Optical Character Recognition) and provide Artificial Intelligence to automate learning procedures.  Solutions based on templates will be time consuming to implement on the front end.
• IT organizations must evaluate solutions that move away from proprietary code and offer a platform that allows analysts to model and design engineers flexibility to deploy.
• BPEL (Business Process Execution Language) provide flexibility to implement long running processes that require both synchronous and asynchronous messaging.