BusinessDay: Machine learning, automation and common sense

• Read original article here

In an era of rapidly increasing competition, business leaders are looking to technology to streamline their operations to deliver products and services faster while lowering costs. Modern technologies such as robotic process automation (RPA) are playing a great part in these initiatives.

These plans are often led by the technology department, with little business acumen, understanding or impact. Executives are asking whether their investments in the automation teams and technology platforms are truly affecting the bottom line.

For the most part, it seems that the focus is on automating tasks — normally those from the back-office that are repetitive and of low value — with little consideration of the end-to-end value streams.

There are many things that technology cannot do, and by relying on a technology-first approach many of these automation plans fail to deliver the expected value. Over the past few years, I have been working with customers as a management consultant on their automation plans. I have seen how common sense is rarely part of their plans and how efficiency and effectiveness gains through automation are rarely achieved.

I consulted with a large insurance provider that wanted to automate its procure-to-pay processes. These were paperwork-intensive with many people involved to process the purchase orders and vendor invoice payments. My client received about 400 invoices per day from nearly 3,000 vendors.

At the end of an initial investigation, it was clear that the team had no overall view of vendor payments. In the chaos of receiving, capturing and processing invoices there was little time to consider the terms in the master vendor agreements. The goal was simply to verify the invoices against purchase orders and then walk over to the various buildings to ask for the relevant signatures.

These processes are easy to automate through a cocktail of technologies such as automated workflows, digital document capturing, machine learning and RPA. The team was able to work faster and the number of errors was reduced. However, speed and accuracy — as valuable as these are — did not deliver value to the balance sheet as the executives had hoped.

Upon analysing the data with machine learning models we discovered an interesting fact. About 10% of vendor payments accounted for nearly 80% of spend. My client had early payment discounts built into most of these contracts, but in the chaos of shuffling paperwork around no-one had much time to take these into account.

We then built a dynamically and real-time updated dashboard to assist the payment team. Our data showed that if the top vendors were paid one day earlier every month it would trigger settlement discounts resulting in a total annual saving of R23m.

Needless to say, this had a great effect on the balance sheet. It was common sense, aided by technology that resulted in a significant business impact. Simply automating tasks would never have achieved this, though it was a start.

My advice to clients is that they should identify the high-impact business areas first. Look for the right technology platforms, but never be fooled into thinking that the technology by itself will result in the goals you have set. Empower your people to take a step back, and rethink their ways of working before they embrace technology.