Is Artificial Intelligence (AI) going to replace Robotic Process Automation (RPA) eventually?

Whenever we talk about Automation; RPA is considered as a low-hanging fruit only for immediate consumption. It is also assumed by many that Artificial Intelligence (AI) will eventually take over from RPA. If we add ‘machine learning’ into this mix and then the confusion further peaks-up. In some cases customers have even asked me if RPA is a right starting point in their ‘automation’ journey!

Perhaps, this is because of the loose definitions of all these terms or because of the way the ‘automation spectrum’ is generally presented. At one point, I myself was trying to alienate from AI as AI was on its path to kill RPA - my immediate focus areaJ But over the last few months; I have tried to address my anxiety and I am sharing my understanding herewith.

Artificial Intelligence: Simplest yet meaningful definition of AI that I came across is ‘a computer program that does something smart’. It could be a pile of ‘if-then’ statements used in a chess program or a complex algorithm used to drive a self-driven car. With this AI definition everything appears simple!! We can say RPA is also AI and Machine learning is also AI. Yet we all know that AI is bigger & better than these two; as it theoretically can go up to replacing humans. [Off course chances of it happening are very slim in the foreseeable future and that brings another interesting term – AAI (Applied AI) to forefront which is explained very well here.]    

Machine learning: Machine learning as the names suggests is a self-optimizing algorithm i.e. a computer program usually designed to maximize the likelihood of its success by auto tuning/correcting based on its own previous attempts. If you tune them right, they minimize their errors by guessing and guessing and guessing again. As per our definition this being a ‘smart computer program’ can very well be counted as AI.

RPA: By now it’s a well-established fact that RPA is a ‘smart’ computer program but only for ‘rule based’ activities and is constrained with indecisiveness. But the question that is still open is, if RPA (basic AI) is a temporary phenomenon to be replaced by something else? (more matured AI perhaps??) And my answer is big “NO” for two simple reasons;

1. More than 50% of the current enterprise processes are simple enough and don’t need any additional intelligence power than what is being offered by current RPA tools & techniques for their automation.
2. For much of the remaining processes; both RPA & AI can in-fact be used together in a complementing fashion. Integration of NLP (Natural Language Processing) & RPA or integration of Chatbots & RPA are few classic examples. More generic example would be – Use AI to structure the unstructured data at the beginning of the process, followed by RPA to process the transactions thereafter. (Check one such recent development here)

In short RPA is here to stay. Important point however is to choose right processes for automation and to choose right combination of tools & techniques for automation! This again highlight the fact that ‘Automation Design” is THE most crucial activity in your automation journey.

I would really appreciate your inputs & comments in this learning journey!