Robotic Process Automation (RPA) tools currently reside at the “Peak of Inflated Expectations” in the Gartner Hype Cycle for Artificial Intelligence, 2018.
As always with technology, during the discovery cycle everyone is tempted to think that it will solve all problems, hence the next stage of the Gartner hype cycle — disillusionment.
There are two ways to use tools & technology:
- You have a hammer, you see a nail, you hit it and it works.
- You have a hammer, you see a screw… you hit it and it “kind of” works.
Indeed, when you have a hammer, everything looks like a nail. In order to avoid inadequate usage of technology, I will try to explain what RPA is as well as its appropriate scope of use.
RPA is a program or a set of programs, that have been developed (usually using a “macro” language) to manipulate other programs. This program runs on a desktop and performs the operation (the macro) as if it were a user, hence why it is often named a bot. For example, it can open your email, grab some text inside an email and copy the text into your CRM or ERP user interface.
Usually with macro-languages, a macro-recorder helps you “implement” the macro. If you want to see how it works, try it with Excel: Search “recording a macro” in Google and it will propose dozens of video tutorials.
How is RPA different from a macro in Excel?
According to Wikipedia: “RPA tools differ from such systems including features that allow data to be handled in and between multiple applications.”
Simply formulated, it is a macro tool that can work with several applications, connecting the dots in manual processes. This is a great evolution; however, you may agree that this proposal alone isn’t mind-blowing.
So why the hype?
The claim (which causes the hype) is the following: RPA systems develop the action list by watching the user perform that task in the application’s graphical user interface (GUI) and then perform the automation by repeating those tasks directly in the GUI. This can lower the barrier to use of automation in products that might not otherwise feature application programming interface (APIs) for this purpose.
It is a way to automate manual processes and save both time and money, which obviously is very appealing. However, the sentence “that might not otherwise feature APIs for this purpose” is important. This sentence begins to answer the question about the scope of use of RPA. If you have an API you should use it, it’s much more reliable and maintainable than a robot.
But what if you don’t have an API? Let’s say your customers or vendors are forcing you to log into their homemade portal to download orders or invoices. This may be a good use case for RPA. If you need to log in every morning into 30 portals and check if there are orders or invoices, there might be a good ROI (Return On Investment) in automating this task with RPA.
Consider the cost of training your customer service team to use these 30 portals, the risk of losing an order, and a simple ROI calculation will give you an answer. However, you also need to take into consideration the fact that the portal user-interface may change. Be prepared for maintenance, as while APIs typically guarantee upward compatibility, it is not the case for a GUI.
Now let’s look closer at the first part of the claim and what is really causing the hype: “RPA systems develop the action list by watching the user perform that task in the application’s graphical user interface (GUI) and then perform the automation by repeating those tasks directly in the GUI.”
This is a more compelling vision. With Artificial Intelligence (AI), a bot just observes what a user does and learns by itself what needs to be done and how to automate even a complex process. As Gartner says: “When choosing a vendor, also ask for future AI-based options.”
RPA is an advanced macro tool that automates simple tasks but to get to the next step, the real thing resides within AI — hence why many vendors use a bit of AI and start calling the technology “cognitive RPA.” This level of cognition is what influences the TCO. If you have to maintain hundreds of bots, what will be the maintenance cost over one year, two years, etc.?
When you select a vendor, here is a list of questions that are relevant:
- What is the level of “cognition”?
- How intelligent is the bot?
- What exactly does it learn?
- Does it learn only once?
- Does it need to be taught?
- Does it continue to learn after it has been implemented?
- What happens if something changes?
- What is the TCO?
- Do I need IT resources?
- Do I need developers?
- For how long?
In Esker’s Order Management solution, we use machine learning to study how users extract data from sales orders. The machine-learning engine identifies where data is located, its format and whether it is always in a fixed place or moving up and down (like a floating total for example).
The engine learns continuously: if something changes, it will learn the change. This AI-driven technology is part of the application, it does not need any IT resources to maintain it, it does not need developers, it just learns by itself.
Now that you better understand this technology and what it is able to do, how will you leverage it in your business processes?