Robotic Process Automation

How to make automated unstructured processes with RPA

Robotic process automation is the technology used for automating business processes. RPA software bots can communicate non-intrusively with your applications and they are easy to implement in any system. You can train them to do the boring repetitive work for you. They mimic human actions but unlike humans, they don't need a break, sleep, or vacation days.

  • Better accuracy,
  • improved productivity and
  • cost savings

are some of the business benefits you will experience using RPA and the reason why we think that RPA is a need-to-have and not a nice-to-have in the modern business world.


The downside

This all sounds great but there are still some requirements a business process needs to meet in order to be a good candidate for automation. A process needs to have some kind of inherent structure for a robot to be able to correctly carry out all the needed work. For example, if we want from our robot to extract some particular words from the document, like name or ID number, we need to put the words in a certain form, otherwise, our robot won't be able to distinct name word from other words and ID number from other numbers in the text.


Looking for alternative

However, there are still some detours we could use if we want to take advantage of RPA and our process is not completely structured.

In one of our showcases, a car leasing company, our client, receives over a thousand requests via free-form email for contract buy-out, which took three full-time employees to read through, prepare the buy-out proposal and send it back to the client.

We wanted to automate that process via RPA, but things were not that simple – the emails they were receiving were all but structured. Every person arbitrarily composed the email and some of them had some data missing.

Using some advanced email parsing, we were able to automate only 60% of the workload. The other 40% we handled by implementing an auto-reply email with a link to the web form in our robot to send it back to the client asking him to fill the web form with the data that was missing. By adding this extension we increased the efficiency of our robot to almost 100 percent.



Fuzzy search for better effectiveness

In another showcase, one of our clients, a bank, receives a lot of requests for debt sign-offs in case of death. The input data is a repository of PDFs containing data in different forms. They need to find and extract necessary data, fill it in application from the bank, search for potential debt, and sign it off if a person is dead. A rather morbid task but our robot had no complaints.

Automating this process was a challenge because input data was unstructured again. We solved this by implementing a fuzzy search in our algorithm. Fuzzy search is a search algorithm that matches an approximate value of the target item. Meaning if we are searching for the word 'robot' it will also find 'robots', 'robut', 'rpbot', and similar 'words'. With the fuzzy search, we took into account possible OCR errors caused by bad quality or a lack of structure in the data which made our robot more effective. If keywords for debt sign-off are found, the robot processes the debt in the banking app and finishes the document. Otherwise, it sends an email notification to the business user saying that the document needs to be processed manually.

Before implementing the fuzzy search our robot was able to implement about 70% of the workload and with its implementation, the efficiency was improved to almost 100%.


The takeaway

Here we see that you can also benefit from the use of partial automation. If it isn't possible to automate the whole process, you automate just the part of the process that has a structure in it and, for example, send the email to the customer to manually do the unstructured part.

Also in cases when you think your robot could do mistakes because of the lack of structure you can implement sending an Excel report back to the user to check whether the results are meeting the requirements, and if not, the procedure can be repeated.


The future is automated

RPA has proven a useful tool in many businesses, and innovation in things like document understanding and computer vision make it more useful every day. However, there are inherent limits to which you can automate processes via RPA, but there are also ways to extend those boundaries. A 60% workload reduction can still pay off, even though 40% of the work will still be done manually.