Technology invades and surrounds us wherever we go. And the business environment is no stranger to it. But sometimes it can be overwhelming and the amount of jargon, concepts and ideas with which it arrives to us can generate doubts or rejection of its adoption.
One of the areas in companies where great efforts are focused is in automation with the aim of increasing the efficiency of their processes. And this is where a multitude of concepts begin to invade us: OCRs, RPAs, Artificial Intelligence, Machine Learning, Cognitive or Intelligent Automation… What does each of them mean? Are they competing or do they complement each other? Which technology is the most suitable for each case?
Well, the first thing we need to be clear about is the goal of all these technologies: to have a machine that performs tasks with no added value that a human person could be doing. This may happen because of an improvement in the efficiency of the process or because there are tasks that people carry out without adding any value to the company. Each process is made up of different tasks, some more complex and others less, but the interesting thing is to understand the whole in order to differentiate these parts. And this is where we will begin to see the differences between these technologies.
Understanding each technology
It will be easy to understand its use cases and possibilities if we understand what is, and what is not, each of the technologies or concepts we were talking about. And we will focus on defining them within the scope of document processing, where Serimag has the expertise and provides a solution that we will discuss later.
On the one hand we have the OCR (Optical Character Recognition). A software that has been with us for years and has no other objective than to convert images with text to a plain text file with the characters recognized within the document. It does not apply any kind of intelligence or comprehension. At best, it imitates the layout of the document in the new digital file. We will not go into more detail during this article on this technology, because it is simply relegated to a small part of the process that an RPA or an IA will carry out: the transformation from image to text.
On the other hand, RPA (Robotic Process Automation) is also a software designed to mimic certain actions that humans perform. They are highly repetitive processes, simple by nature and with a very basic decision level and always based on rules. In the cases we are dealing with, they imitate the actions that a person would have in front of a graphical interface of a computer with a keyboard and a mouse. An example could be an RPA that opens a web page, selects a drop-down option, downloads a file, renames it to the current date and saves it in a repository according to a defined criteria.
And finally we find Artificial Intelligence and autonomous learning. The first includes an endless number of different technologies, such as Computer Vision, Natural Language Processing or Machine Learning. But all of them seek to imitate the behaviour and judgment of a person, the generation of hypotheses and the analysis to respond to known cases or to be able to act in front of new situations. Therefore, we understand that it is a step forward in process automation because it is not based on predefined rules, it can take part on more complex tasks and it will know how to manage exceptions better. A clear example in the document sector is the extraction of data from unstructured documents, where there is no basic rule for their location. Or the management of incomplete, modified or trimmed documents that make a correct recognition of data difficult
Differences and convergence between them
Once each technology is understood, let’s see how they coexist or should coexist within the business environment. Here it becomes an essential part to identify and understand the process to be automated, in order to understand each of the parts that make it up and treat them differently.
The RPA is a tool with a very fast development and implementation and that works with the environments and applications already existing in the company. Identifying these processes and applying these robots will bring quick benefits. If an Artificial Intelligence were to be incorporated here, it would imply much higher costs and times in order not to result in any benefit.
On the other hand, implementing an RPA for the comprehension of a document seems rather titanic at the level of effort: there will be so many rules to be implemented and so many exceptions to be contemplated that its development will be prolonged and, something more important, its maintenance will make it unviable. Here is where an AI makes possible the project with a lower development and maintenance comparing with the good results obtained.
But let’s continue with the example mentioned above: where the RPA connected to that website to download a document and save it according to some criteria. Let’s imagine that this criterion is the identification of the type of the document: if it is type A, it must be saved in folder A, or if it is type B, in folder B. Who should carry out this analysis? If we do it with the RPA we can go with the complexities mentioned before. However, the joint work of the RPA sending this document to the IA to emit a verdict so that the RPA can follow its flow seems to be the best option.
Understand the process for choosing technology
And that should be the process followed by every company with every process it tackles. Understanding the process in its entirety, each of its parts and know how to differentiate the advantages and disadvantages of each technology in order to apply as appropriate in each case. Companies today produce a huge amount of data, but about 80% of them are complex or difficult to process by systems based on rules such as RPA.
The goal is to get the “digital worker”, the robot-software, to free them from these repetitive tasks and thus be able to focus ourselves to what we do best: think creatively. Cognitive Automation (or Intelligent Automation) should seek to unify these technologies to take advantage of the best of each of them: allow RPA to imitate people’s actions, and leave it up to IA to imitate our behavior or judgment.
A connection that is already real in serimag
At Serimag we have been working for years and developing our own Artificial Intelligence solution for the automatic processing of documents. And in some cases, we have been required to be part of a complex process which incorporated RPA system to control the flow of documents.
As an example, we have applied our solution in the banking sector with mortgage processes, in which information provided by the client must be verified with information downloaded from the Land Registry. Here, both technologies work together: the RPA is in charge of obtaining this documentation from the Registry’s website, it is sent to our TAAD automatic document processing module, awaits its verdict for the extraction of certain fields and then acts according to its comparison with the data provided by the client.
An effective work-force, contained costs, low impact on implementation and, above all, focused on obtaining results.
