Manufacturing organizations have often been on the forefront of innovation. Artificial Intelligence can go beyond making machines smart and into making crucial support functions like invoice management highly efficient.

The manufacturing industry in India is riding the crest of an upswing in demand from a wide range of user industries. This demand is driven by the burgeoning spends on infrastructure as a result of the booming economy. Such demand-driven success notwithstanding, manufacturing organizations need to continuously improve their processes and operations through innovation. Success in a manufacturing scenario as the Japanese showed the world is not just about sales and marketing, but also about operational efficiency, productivity, speed and turn-around times, customer service and supplier relationship management.

When it comes to building efficiency and productivity into their systems and processes, manufacturing organizations need to look beyond the supply chain. There are a range of supporting functions like procurement, transportation logistics, finance, accounting and taxation, that are critical and integral to the manufacturing organization’s sustainable profitability.

Case in point is invoice processing, a tedious, manual activity prone with errors, multiple layers of approvals, with scope for fraud at each level. A manufacturing organization’s operations face the danger of being affected due to glitches in the invoice processing function. Suppliers may delay or withhold crucial parts affecting the entire operations. Errors in quantity or quality shipped or the price of the goods, inadequate approvals, invoice disputes due to wrong credits and payments, invoices submitted without purchase orders, cost center errors, duplicates, errors in purchase order allocation and mis-matches in freight, are but a few of the invoice management issues. The effort and time required to remedy errors in invoice processing are huge and disproportionate.

Artificial Intelligence: A New Paradigm

Artificial intelligence used in the processing of information together with a web-based workflow system is an innovative method of minimizing invoice management problems, and uses the best of tools available for a manufacturer. From the standpoint of improving efficiency and savings costs, it can have incremental benefits. In some cases where thousands of invoices are processed daily, cost savings are as much as 50%, manpower reduction from 10 to 2, and time for processing down from the original two weeks to the much-improved 2-3 days. This innovation also has the added advantage of integrating with the existing enterprise applications like SAP, SCM and CRM.

Artificial Intelligence (AI) used for processing information on documents is a divergent view from a robot environment that manufacturing units typically envision. Machine learning have mostly been used in CAD / CAM tools to improve architecture or reduce design costs. ‘Smart’ in manufacturing has been limited to intelligent modules that communicate with each other and collaborate.

The beauty of artificial intelligence is that it recognizes patterns, differentiates and analyzes information. It can be trained and configured to meet the users’ needs. In the context of processing information, it enables a PC to do whatever humans do with paper, using natural language processing. This is ideal for processing any type of documents, whether un-structured, semi-structured or structured. Invoices typically fall under the category of semi-structured or un-structured documents, which do not follow a standard template, and thus increase substantially the scope for errors if done manually.

To be effective, the AI solution need to have business rules built-in, uses fuzzy logic and image topography analytics. It needs multiple validations for data and fields for seamless validation of output data based on in-built rules & customized set of validations, and a confidence level. A fully-automated invoice processing and management solution may not suit the needs of many manufacturing organizations. The ‘confidence level’ thus acts as a provision for user intervention, which is asked for if the confidence level for a particular field is below the accepted limit.

An information processing solution needs in-built automation that enables intelligent rendering of conventional pre-processing operations like de-speckle, de-skew, rotation and de-size. Through micro-registration of structured forms (for instance, boxes), the solution should be able to align such forms and identify them to standard orientation if they lose alignment while scanning. It needs to identify and restore document integrity. This means that if the capture loads have instances where the documents have irregular page sequences or page numbering errors, the solution should recognize these documents and re-sequence the pages / images (if needed), thereby establishing document integrity.

In the case of the intelligent web-based workflow system into which the capture and processed information enters, it should have an automatic routing or manual routing of invoices for approval. This really depends on such facts as the user, the type of invoice, amount of invoice and the speed at which it should be processed. Invoices in this system should be capable of being tracked at any point in the approval process, and it should also allow status change and reporting in real-time. An added feature can be access of the invoice image online with notes attached to the invoice. The web-based system goes a long way in improving supplier relationships as suppliers can submit invoices online, view their status and track remittances using the supplier portal. The entire system, if available in a dashboard with real-time information and integrated with other enterprise applications, can act as wonderdrug for a manufacturing organization’s critical invoice management.

Challenges in Traditional Invoice Processing

Invoice processing has traditionally been a paper-intensive process. It goes through multiple levels of approvals and hand-offs and increases the probability of injection of manual errors. Each vendor / supplier submits invoices of different shapes, sizes and formats (unstructured documents). It becomes extremely difficult for the processing team to adapt themselves to the necessary fields like Invoice Amount, Invoice Date and Vendor Name as the position of these fields are at different places in different invoices. Some of challenges faced during processing paper based invoices are variance in invoice formats (non-uniformity in field position) and capturing invoice information and its storage.

In summary, a invoice management to work, it needs to meet the following challenges.

Volume: Other challenges in the existing invoice processing process for many manufacturers include the volume of documents. An organization receives thousands of invoices, in some cases millions, annually. Such large volumes of documents to be processed hits the efficiency and productivity of the processing team. The cost an organizations incurs in correcting manual errors is compounded with lose of early payment discounts, or conversely late payment penalties.

Multiple Sources: Invoices are received at an organization through multiple sources and the task of integrating the flow of the invoices is daunting. An invoice can come through emails, through courier, fax, web-based forms and other electronic formats. The multiple sources adds to the processing time.

Document Management: The paper-intensive nature poses immense challenges not only in terms of its processing but also in terms of its storage and filing for further processing and future references. Conservative estimates by various research agencies puts the amount of paper used every day at 700 pages per person, per year.

Compliance Challenges:

An organization must have an efficient and structured process in place to handle vendor invoices. With regulatory compliance requiring exercising higher control and VAT / Income Tax assessment getting more difficult, compliance challenges are increasing.Exception Handling:

An inefficient Accounts Payables process can pose serious challenges while processing an invoice if it doesn’t have processes defined for handling exceptions. Some scenarios could be an invoice without a purchase order, general ledger code missing for a given line item or a missing vendor name.

Updated Information: Updation of information related to existing and new vendors is a challenge and delays the process before getting approvals for processing the invoices into the next step. There are similar problems with the Ledger and Sub-Ledger updation too.

Integration with Applications: Invoice Management needs to be integrated with different applications such as Procurement Management System, Treasury System and Corporate Management Information System.

The extent of accuracy and efficiency that a solution with such features provides is manifest in the lower effort spent in resolving problems in invoices management.

In a manufacturing scenario, the results are translated to improvement in process efficiency by over 50%, and an accuracy of 99.9%. If one includes logistics and transportation, then such a solution improves process efficiency by 93% and delivers an accuracy of 99.99%! An important point to remember while choosing a vendor to implement such a solution is its domain expertise in manufacturing combined with the depth of technology deployment experience. Such a solution is sure to give the cutting edge to a manufacturer in an increasingly competitive world, and aid in building the factory of the future.

This article appeared in the October 2009 issue of "Industry 2.0 - Technology Management for Decision-makers". www.industry20.com

Please send feedback on the above article to business@datamatics.com