Charles Howard Duell of the U.S. patent office famously stated in 1899 that “Everything that can be invented, has been invented.” While this notion clearly isn’t accurate, we can all look at many of our own internal procure-to-pay (P2P) processes and see that little has changed over the past decades. Procurement has long looked at innovation with an “If it ain’t broke, don’t fix it” approach. In the age of the intelligent enterprise, many are revisiting this sort of thinking. If we look at some of the steps in our P2P process, maybe we can improve them.
A straightforward process?
In its simplest form, procure-to-pay is pretty straightforward. First, we either generate a purchase requisition or request an item out of a catalog. The requisition goes to a purchasing agent who: performs the search for the item, generates a request for a quotation, or simply approves or denies the requisition. For approved requisitions, vendors are identified and purchase orders are cut. Once the goods are received, an invoice is generated and sent to the purchasing entity. Invoices are matched to shipping notices; the order is validated and released for payment. The terms of payment are checked (to ensure discounts are properly taken), and the invoice is sent to accounts payable (AP) for settlement.
This sounds simple, but for most AP and purchasing teams, it’s anything but. How do they process free-text purchase requisitions versus those that enter through the system? What happens when a product or vendor is not on the vendor or product master? How do they make strategic sourcing decisions? What happens when human interaction is required? What about the payment system? Who checks if discounts are taken? What if there are outstanding disputes with a vendor, indicating a block on a certain payment? How do they identify procurement fraud? Our colleagues in the supply chain and in the payments processes routinely face all of these questions (and probably thousands more).
Why can’t repetitive P2P processes be automated?
But if they are repetitive processes – and we see the same types of questions all the time – why aren’t we automating them? Why aren’t we training our systems to capture these conditions and the actions that we take to remediate the exceptions? It can’t be a lack of data. In any given large organization, purchasing stats are readily available. In fact, most large organizations have already invested significantly in their spend analytics programs.
Analysts have estimated that organizations that do not use spend analytics miss out on 10%-15% in savings in unmanaged spend categories and another six percent on managed categories. And that is on data that has been recorded (post-purchase or retrospective data). What if you could predict where losses would come from? What if you had systems that were actively learning from your data to provide statistics and predictions based on your own data? What if your systems “knew” how to process repetitive tasks to free up your purchasing and payables team to track down emerging one-off problems and money-saving opportunities? This is the opportunity for machine learning and the intelligent enterprise in P2P.
By using data science and automation against large sets of purchasing data, you have opportunities to:
- Reduce free-text item purchases – Identify what, when, to whom, and from whom free-text purchases are likely to be made and develop a strategy to remediate these practices.
- Automate sources of supply assignments – More tightly control and monitor these assignments in real time, and use machine learning algorithms to anticipate them, to negotiate better contracts with vendors, and to ensure consistent performance in your organization.
- Manage cash discounts – Identify reasons why discounts are not taken (for example, is it specific to certain agents, products, or seasons?), build programs to create alerts, and train employees to help your organization to be better cash managers.
- Automate manual processes
Why not harness the power of the intelligent enterprise?
It is vitally important to see how the intelligent enterprise not only helps manage retrospective analytics and statistics, but does so in a prospective manner using predictive analytics, as well. Unlike Charles Duell of the U.S. patent office, we now know that our P2P processes are constantly inventing themselves. It’s up to us to harness the power of the intelligent enterprise to enable it and to empower our employees to constantly reinvent the organization. We can do this by developing the building blocks to improve the working lives of individuals and businesses as a whole.
If you are interested in co-innovation projects to jointly advance these topics, please let us know.