Using AI to Detect Supply Chain Fraud

Editor’s Note: Norman Katz, president of supply chain consultancy Katzscan Inc., writes a monthly column for Supply Chain Management Review. Katz’s column appears on the third Monday of each month.

People tend to get overwhelmed by all the emails they receive highlighting potential or real problems. All that stimulus tends to have the effect of turning people away from actually looking at what they really should be looking at, and letting problems slip through the cracks. Part of the issue is discerning whether a warning is really a problem, and whether an error is innocuous. This is dependent upon the programming in the software that is determining the severity of the condition and sending the message.

I’ve been faced by this myself numerous times, especially on consulting assignments when I have been tempted to turn off or shift the message signaling that EDI or ERP systems generate. But I have been reluctant out of the concern that my clients might miss something important. So, I help them establish mail (e.g., Outlook) rules to divide-and-conquer the influx of messages to better manage the communications. It might not be a great solution, but it’s better than missing a critical message.

Artificial intelligence (AI) goes beyond pattern recognition to attempt to reason as to why the pattern occurred. I’m not going to write about whether AI is spot-on or how objective or subjective it is in its current state. I’m just saying that between advanced reporting such as BI (business intelligence), the capabilities of ML (machine learning) and AI, there exists sophisticated software systems to pick out patterns and perhaps even offer a possible reason why. And with a person involved, the technology-people pair is a powerful combination.

Detecting fraud in supply chain operations can be done by using the same supply chain systems – ERP (enterprise resource planning), EDI (electronic data interchange), and barcode labeling and scanning – and the transactions that they output that are used to measure supply chain performance. The same systems, the same metrics, just a different perspective. What a great return on investment this is.

Auditing just a single system to check for fraud isn’t sufficient enough depending upon the systems involved and their integrations. For example, where there is an ERP-EDI integrated relationship with EDI850 purchase orders inbound from customers, the source of truth for purchase orders would really be in the EDI system, not in the ERP system. Conversely, in this scenario, for outbound EDI810 invoices from the vendor (the customer), the source of truth for the invoices would be in the ERP system, not the EDI system.

Is there shady activity going on in the warehouse or distribution center? Cross checking received goods against the purchase order header and detail at the point of receiving would determine this. Too many or too few received against the purchase order could be a problem, backorders aside and should be taken into consideration. And trying to receive an item not originally on the purchase order: that should definitely be sending off a “look-at-me” signal.

Are the shipment weights and volumes within tolerance of what they should be? If your item master data is clean and your scales are calibrated, then there should be no short-shipping or over-shipping. Use the EDI856 advance ship notice and the physical receipt to ensure that what you were supposed to receive is what you actually received.

Before you pay an EDI810 invoice, ensure that you are paying for what you received, and received in first-quality condition. Ideally, there would be a trusted vendor/supplier relationship that would alleviate the need for inbound goods inspections. But if inbound goods inspections are required, consider integrating your quality inspection process and system to record the results and essentially pay (fairly) on performance.

Auditing the general ledger for problems is after the fact and too late. Illicit activities need to be caught at the point of the infraction so they don’t ripple through your supply chains and manifest themselves into something worse. Supply chain fraud is an accounting problem after the fact: when there are penalties to pay, chargebacks, damages due to lawsuits, and costs due to missed advertising opportunities, corrective communications, and recalls.

Detecting and reducing supply chain fraud can be done using existing supply chain systems and the familiar transactions that they produce. As important as it is to protect your company from outside attacks, don’t forget to do due diligence from insider threats as well. With the advanced software technologies available on the marketplace, assessing potential illicit activity within your supply chains is now more achievable than ever. In examining your supply chain activities, you might be surprised to find how much is really within your control to examine, even if you outsource different aspects of it.

Whether differentiating a warning from an error, or distinguishing poor performance from illicit behavior, partnering the right software with people and enhancing the subjective perspective helps to add a layer of due diligence that can reduce the manual efforts, improve operational efficiency, and increase the odds of catching a problem before it infects and negatively affects your business.    

About the author:

Norman Katz is president of Katzscan Inc. a supply chain technology and operations consultancy that specializes in vendor compliance, ERP, EDI, and barcode applications. Norman is the author of “Detecting and Reducing Supply Chain Fraud” (Gower/Routledge, 2012), “Successful Supply Chain Vendor Compliance” (Gower/Routledge, 2016), and “Attack, Parry, Riposte: A Fencer’s Guide To Better Business Execution” (Austin Macauley, 2020). Norman is a U.S. national and international speaker and article writer, and a foil and saber fencer and fencing instructor.

SC
MR