Can Robotic Process Automation fix IT ?

When petroleum jelly was discovered on an oil rig & it’s amazing healing properties, people assumed they have found a miracle elixir. They assumed it can fix anything you apply it to. Numerous attempts to fix leather shoes, broken china and 'God knows what' were made. I guess at some point in time, there was a realization of specific utilities where petroleum jelly was phenomenally good & several other functions where it’s just a proxy of something much more efficient and robust. With Robotic Process Automation (RPA), not everyone has reached this realization... Yet!

In my initial exposure and since working on RPA projects; primarily providing Business Consulting and executing the Business Project Manager role; I was lucky to see the story unfold from a different perspective. I want to share my take & some RPA traps you must avoid.

But before that, why is RPA buzzing?

Automation was taken up as a serious agenda across several industries & all are at different levels of maturity. Now with the majority of service industry putting Automation agenda in their daily driver; visibility has skyrocketed. Your bank most likely has some automation that you experienced.

A simple google search will expose you to terms like Machine Learning (ML), ChatBots, AI, RPA, etc. I don’t want to define & explain them here, but let us look at the environment we are in today.

Customers are Aware

If you have not caught the Automation bug, your customer will infect you soon enough. They are aware of it & want you to use it for their advantage; Whatever the advantage maybe; Cost, Quality, Response time you name it. You simply cannot ignore it and in high likelihood, an Automation KPI will be part of the monthly dashboard customer demand soon enough. But figuring out what needs to be automated is your job, so what will you do?

Your Options Are...

AI/ML – You need a focused problem that you want to solve, a serious talent pool for running the effort & lots of data and people who understand the data, to succeed in a machine learning or Artificial Intelligence endeavor. The technology is mostly opensource, you need to put together an infrastructure to run it on (bigger the infra, faster the results). However, results might not be useful to get-go. Iterations are the name of the game. Surely you won’t be able to cook up something before the next big customer meet. This needs time and refinement.

ChatBots – I will stick my neck out and say this is the gimmick that you use and try to lure people into the illusion of Automation. Any ChatBot solution you implement, straight out of the box, it will be great in conversations. However, this is a monster that is always hungry for data and patterns. Essentially part of ML, the ChatBot will make sense if you have well-structured data and a team (larger the team, quicker the results) to constantly train (Test and Correct) how the BOT responds to the questions posed by impatient humans. After all this effort; adoption becomes the biggest challenge. Customer is represented by a small set of people (touchpoints) & if they get irked by any inaccuracies or underwhelming performance by your ChatBot; they will stop using it & that is the end of all your efforts, till you convince them to use it again!

RPA (Robotic Process Automation) – Imagine this; what if you can record a repeated action and run it on a loop with your triggers. You receive a Sales Order - Trigger a Purchase Order; you receive invoices – trigger matching process and issue payments; there is a KPI degradation – Trigger alert emails; Lots and lots of possibilities, or are there?

In my personal opinion, RPA is relatively quick and with some TLC, you can derive value easily. This article is more focused on RPA, hence I would like to explore this a bit more. What is RPA can be answered by a google search, but if you are about to start an RPA program, Let me share a few trappings which you must avoid.

• FIX it First: Don’t Automate a process without really looking at areas of improvement first. Once you automate, significant human involvement will go away. Several processes in an organization “looks” good only because of the human element. They are the sticky glue that holds it together & trust me you don’t want a BOT to tell you the process is broken.
• Make Room for the BOTs: A lot is assumed in a process document, but with a BOT in play, there have to be clear definitions of handover between processes. A BOT will also generate its reports and I recommend spending time defining them well. A human next in the process chain cannot call and ask a BOT what’s up! BOT actions need to be report-able and audit-able by downstream or upstream folks. The BOT will take a slot in the process chain, make room & don’t ignore it.
• Need more people – Yes, I mean it... There will be new roles needed to upkeep the BOT, fix errors and keep it in tune with the changing landscape of the business. Technical & business folks will have to put efforts to keep it working. The net reduction in your workforce will be there, depending on what you are automating using RPA. Get ready with a talent pool to handle these new roles, hopefully, some of the same folks you are going to cut because of automation efficiencies!
• Commit – Don’t half-ass it and keep the manual process running parallel to the BOT. Because one Monday morning the BOT will break; Yes.. your process will be put to a standstill & if you always keep the safety net around; people will treat this as an experiment and you will never transition to an RPA optimized operation. Use the talent pool to keep it running, refining it and add more processes as you go along.

As a bonus, I would also like to add the trappings of RPA platform marketing, where they present their platform as the "Swiss Army Knife" of automation. Most of the features an RPA platform offers sounds great; but trust me, a lot of it will prove useless for you. Avoid the trap to use RPA platform components just because you can deploy it and its available. Automation should happen where it’s needed and useful; not by what is possible.

Hopefully, my observations will help you to avoid making RPA the digital petroleum jelly and mindlessly apply it to every crack in the process; that is a costly way to discover the area of maximum value.

! Best of Luck!

P.S. Share your thoughts and opinions in the comment section. What other trappings do you see, where do you disagree? Thank you for your attention.