Robotic Process Automation: Which bots are for your business?
Robots are a unit of automation. On the manufacturing floor and in warehouses, industrial robots have transformed the landscape. Now it is time for software robots to optimize business processes. There are various types of robots, each referred to with different names depending on the vendor of choice. But rather than focus on actual names, let’s focus on the concepts behind software bots utilized in RPA.
In the concept of front-office bots, referred to as attended or robotic desktop automation (RDA), these robots require human intervention and typically run from a user’s desktop or laptop. You will find these robots automating manual and repetitive tasks. They take form in examples like virtual assistance, Know Your Customer/Compliance processes and sales or analyst activity. Well-oiled machine atmospheres like call centers tend to have some of the highest front-office RPA adoption rates in recent years.
In opposition to front-office is a back-office RPA concept. Referred to as unattended, back-office RPA does not need human intervention (just like any background or batch process) and is controlled by the server-side controller via a scheduler. They are used to batch a list of the activities that do not need human interaction or process. Back-end processing that often take advantage of RPA are processes such as data validation or verification, reconciliation tasks and financially-related batch activities. Knowing this, it makes sense that ecosystems within human resources and supply chain management benefit from back-office RPA.
In addition, an evolution of RPA bots as cumulated into intelligent bots. Outside of front- and back-office RPA, there are software bots that are based on artificial intelligence (AI) and can intelligently analyze data, both historical and current, to make decisions and predictions based on the logic parameters. A common example of an intelligent bot is a chatbot, AI software that can simulate a conversation (or a chat) with a user in natural language through messaging applications, websites, mobile apps or telephone.
Intelligent bot or not?
When RPA is discussed, sometimes it is assumed that all bots are intelligent, but as we’ve covered front and back-end bots, we know this is not the case. Not all RPA bots need to be intelligent, and a majority of RPA robots that are deployed today are not. The question arises: how do you decide if you need an intelligent or regular bot?
Working at a large service integrator, we often determine which bot is best by using a model of structured and unstructured data coupled with standard or nonstandard data. To better visualize this process, I created a simple tool to support this concept. It starts with knowing the type of data and format the robots are going to interact with.
Structured data is comprised of a high degree of organization that makes data sets easily searchable. For example, databases, spreadsheets, catalogs, legal records and census records are all highly organized and searchable by numbers and categories.
Unstructured data is everything else. In this way, there is no organization of data. It may have its own internal structure but does not conform neatly into a spreadsheet or database. Examples are email messages and word processing documents with various kinds of formats.
In a different spectrum, standard data is character or numeric data values that read as lists, columns, formatted or inputs. Non-standard data is unstructured text, handwritten notes, pictures and audio files.
Now that types of data and formats are known, using the tool is a matter of 3 simple steps.
1. Determine whether the robot will deal with structured or unstructured data.
2. Determine the format of the data, whether it's standard or non-standard.
3. The intersection of step 1 and step 2 will provide the type of bot needed.
Let's say your company processes invoices from various vendors, and the invoice format is different for each vendor. Also, the invoice content varies each time based on the services or products provided (even for the same vendor). If we want to build a robot to automate invoice processing, apply the tool to determine what type of robot is needed in this context.
4. Since we know the type of document is an invoice, the data falls into structured data category (even though the invoice structure is different from vendor to vendor).
5. The data format from vendor to vendor is different, and the data format is unknown for the future vendors. This falls into a nonstandard data category.
6. With the cross-section of structured data and nonstandard data, the context calls for an intelligent robot or regular robot. Although, if you add new vendors often and would like to process invoices automatically, then it makes sense to go for an intelligent robot. If you don’t add new vendors that often, then it makes sense to go with a regular robot. In the case that you need to add a new vendor, you will update the robot rules (or code) manually.
This tool gives you an idea to determine the type of robot; however, the exact type of robot will be determined and confirmed after a detailed analysis of the use case. Based on the outcomes of this, a regular bot or intelligent bot will be selected, and sometimes a combination of both is relevant.
In the next part, we will review how to get started with RPA.