Machine learning + RPA transforming purchase order processing

In all organisations there are processes that are purely the ‘cost of doing business’. Like traditional retailers who still provide checkout staff (yes, an argument can be made at a brand/customer proximity level but…) to process their sales, there are many sales and operational processes that themselves raise no incremental value, but are necessary for the business to run.

Like purchase order processing, or supplier invoice processing.

Businesses want customers to place orders. And they need to pay suppliers. But the handling of orders themselves, or in another case study on supplier documentation management, the act of processing these orders is pure cost.

Formats vary. Errors are made. Time is taken. Cost is incurred.

Our technology solution tackles this challenge head-on, operationalising this process with machine learning and automated processing of either manual operations or unnecessarily expensive ones.

Our client came with the challenge that they receive over 20,000 orders a month from over 900 customers. Whilst some customers happen to use similar platforms, all of the purchase orders have the potential to have a unique layout. Formats vary between customers, and also within customers.

Some purchase orders are sent via company contacts, some via automated flows, and some to mailboxes. Some contain the information that is needed, and sometimes not. The product of this is high variability, high inaccuracy, and high ‘human engagement’. All of this drives inefficiency and risk, but also team frustration, as people are having to manually enter, investigate, and fix issues relating to orders.

The existing solution was a combination of an expensive automation tool, that had high promise of automating the process, but in practice resulted in an unsatisfactorily high error rate. And came with a six-figure annual subscription cost.

We adapted and deployed our custom robotic process automation solution to their order process. It contained the transformation of the platform, with the corresponding evolution of operating processes and team training.

The platform machine reads incoming purchase orders directly from a mailbox, and automatically enters them into SAP (parsed XML). Order origins are identified, sold to/ship to’s matched, and order details logged for authorisation.

Purchase order data is aggregated and displayed in a presentation layer with associated filters/slicers. This allows business leaders and customer success managers to see real-time information on incoming orders.