Skip to main content



AI vs. RPA



Artificial Intelligence (AI) and Robotic Process Automation (RPA) have been on everyone's lips for the past couple of years. Gartner has predicted that AI's impact on the number of jobs might be negative at first but turns positive over time.


In the sense, the argument is at least in the early stages of believing that, in the first stage, the AI applications will be utilized in particular for the evolutionary development of "old" activities like RPA. The use for artificial intelligence are basically limitless, as RPA almost always aims to automate human work. How do these two relate to each other?

RPA is stupid and AI is smart. I mean, at least for the time being, no deeper AI features have been embedded in RPA tools. And why would it be? The point with RPA is to make the tool do the same processes that a human does, while teaching an AI model always requires a certain amount of data and a particular use. It is easy for RPA to implement “if-then-else” -type structures, but it is more difficult to implement large-scale data-based statistical reasoning.

Because RPA imitates human activity on a computer, it would be logical for it to be able to imitate human reasoning more deeply than with simple condition structures. Here, artificial intelligence comes in handy, but the solution is usually implemented outside the RPA software, for example API / Rest interface. Applications for this include, for example, invoicing, classifying requests for support, and even extracting certain information from a handwritten form.

We have implemented an interface at Enfo, to send for example an image file, a pdf document or written text. Behind the interface is a taught model that is used to deduce the desired information from the uploaded material. This information is returned to the inviting party, such as the RPA software, with authenticity probabilities. RPA software, in turn, can take advantage of the return values in its workflow, if the probability of the values is enough. This solution extends the scope of RPA significantly.

The confrontation AI vs. RPA is therefore not really necessary. We could rather put a + sign between the terms. RPA solutions can already be genuinely improved and enhanced by AI, and we will see more of these applications in the future.


Enfo AI classifier


Tommi Salmi is the vice president of Analytics & AI at Enfo