In our previous blog post, we discussed all of the different technologies (including humans!) that can be used to support the capabilities of Robotic Process Automation (RPA). In this piece we want to look at a few new technologies that have the potential to make the process of actually implementing RPA a whole lot easier.
The implementation process for RPA generally follows three stages: identifying the processes that are suitable for automation; mapping the candidate processes to understand what is happening and when; and configuring the RPA software to carry out the mapped processes. Again generally, the first two stages are primarily ‘business analyst’ roles, whilst the third is carried out by ‘RPA developers’. (It should be said that it is possible for the business analyst and developer roles to be done by the same person, and, in some instances, this can be preferable). Most organisations, at least in the early stages of RPA adoption, look to specialist RPA implementers to carry out these activities.
Often, these initial stages can take a long time and are, by their nature, people intensive. So, what are the forward-thinking RPA specialists looking at to help speed up these activities for their clients?
In the first stage, to identify candidate processes for automation, the business analyst consultants will be looking across the organisation for manually-intensive, repeatable processes. Rarely are there readily-available process maps that perfectly describe every process, therefore most of the time identification has to be done through speaking to as many people as possible in the organisation, especially those that actually carry out the work.
A new approach, using artificial intelligence, could significantly speed up the search for those ideal automation processes. In this case, the software uses Natural Language Processing to ‘read’ all of the emails that go back and forth between employees. Generally, where there is email ‘chatter’ there is inefficiency, and so the software (re:infer is a good example) can seek out those processes that are overly-manually or are broken. As a first pass, this is an efficient way to identify those processes which could make candidates for RPA. (The second pass would be carried out by the business analysts to confirm suitability and filter the list down further).
Once the candidate processes have been identified they need to be mapped down to keystroke level. Existing technologies, such as UIPath’s Recorder, do a good job of collecting all the clicks and keystrokes but this is generally managed by the business analyst. Also, the recorders tend to be quite limited in terms of what they can and cannot record.
Using artificial intelligence again, the process mapping activities could, to a large extent, be automated. Cutting-edge machine learning start ups such as Mimica are beginning to create useful software that can sit on a user’s machine and automatically create process maps based on their activities. The big advantage of this approach, apart from recording screenshots of each action and the time stamp, is that all of the exceptions in a process can also be captured every time they happen. The resulting process map shows all of the branches and what percentage of time the process went down each branch. It can also identify where unstructured data is used in the process so that, by putting all of the information together, it can provide a ‘suitability for RPA’ score for each process.
Just like UIPath’s Recorder, each process will still need validation by a business analyst, but it means that the sometime laborious work of process mapping can be significantly accelerated.
These technologies are at the cutting-edge of AI development right now but are starting to be introduced into the real world of RPA implementation. One day you could imagine putting all of them together to identify and map the processes, and even getting the AI system to code the RPA software, but we are a long way off that yet. We certainly still need business analysts and RPA developers, especially for the really intelligent work of how to improve processes before we automate them, as well as the softer issues around change management. As in most of business, the ideal scenario is going to be a combination of technology and people.