Robots + Automation = sounds like a gloomy future. This week we take a look at Robotic Process Automation (RPA), its staggering growth and discuss the great unknown.
Let’s be honest
There are some things at work we don’t like to do. Whether it be expenses, timesheets, formatting slides, work is filled with routine, boring tasks. But what if this work could be automated and taken over by software, allowing you to spend more time doing engaging work?
Robotic Process Automation (RPA) is a relatively new offering that looks to automate routine tasks done by humans. According to the Institute of Electrical and Electronics Engineers (IEEE) Guide for Terms and Concepts in Intelligent Process Automation, Section 2755, which I know you all have at the ready, RPA is a “preconfigured software instance that uses business rules and predefined activity choreography to complete the autonomous execution of a combination of processes.“
The title is confusing; RPA does not actually involve physical robots sitting at desks, doing your work and chatting about this weeks upcoming football games, but instead these automated process sit on virtualized servers in a data center, doing structured tasks once done by humans. The easiest and most basic way to think about RPA is Excel Macros that have been supercharged by computing.
What is an example of a RPA?
This great link lists 30, but I’ll cover invoices. Let’s say your company works with a lot of vendors to do business and you handle a lot of invoices. Instead of having someone read, catalogue and put the invoice data in a database, you setup an RPA process that automatically downloads the invoice, extracts data from it, uploads it to the database and clears the payment.
Sounds pretty boring
That’s the point. RPA is supposed to be taking care of these boring and monotonous tasks and because a computer can do these types of tasks much better than a human, growth of RPA has exploded. One of the leaders in RPA, UiPath, was valued at $100mn in April 2017, $1.1bn in March 2018 and then $3.1 bn in Sep 2018.
Estimates for RPA adoption are also crazy. Morgan Stanley expects an S curve adoption of RPA and expects that 90% of companies will have some sort of RPA by 2026.
But what happens when RPA gets AI/ML abilities?
Here’s where things get complicated. Originally, RPA just tracked out keystrokes and mouse clicks, so there was really nothing “smart” about it. In our invoice example, if all of the invoices were the same, the computer could just copy and paste from the same location, but if the invoices were different, which they almost all are, the process would break down because the computer only knows how to copy/paste from a certain area.
However, with advances in AI/ML, such as image processing and natural language processing, we can make our RPA smarter. Now, even if the invoices are different, we can use AI/ML to extract things like name, company, amount due, making the process much smarter.
That’s just the beginning
As AI/ML becomes more capable, RPA manufactures will integrate more AI/ML features, leading to more things being automated. Companies using RPA solutions are certainly going to be pushing for these features, as according to Morgan Stanley, “FTE (full time employee) savings from RPA tend to start at 10-20% and are 15-25% at the top end rising to 30-50% with AI/cognitive solutions.”
The great unknown
As we start implementing AI/ML into RPA, questions of companies downsizing their workforce and automating jobs emerge. The data so far is positive, and according to research done by Harvard Business School on cases where companies have implemented RPA, few have shown reductions in headcount and many companies have gone so far to promise employees there won’t be layoffs. But these are simple cases of RPA implementation that touch a small amount of the workforce.
At the recent World Economic Forum, there was a much different tone with the NYTimes saying “The corporate elite’s A.I. obsession has been lucrative for firms that specialize in “robotic process automation”, with many of the attendees looking to automate some of their workforce. What they realize is most people have some part of their job that can be automated. McKinsey estimates “in about 60% of occupations, at least one-third of the constituent activities could be automated”, which means that RPA will touch a lot of people.
How employers decided to implement RPA will be crucial
If RPA actually fulfills its promise of allowing people to do more creative and engaging work, it will be a huge positive. However, if companies start cutting jobs because of RPA and don’t offer re-skilling or new opportunities to those who have been replaced by RPA, things could get ugly. Companies deploying RPA will have to think hard about how to frame the introduction of RPA and what they are going to do given the savings. Do they reinvest their savings into re-skilling the workforce or give that money back to shareholders?
Lots of difficult questions for a Saturday morning.