RPA may not be sophisticated or complex, but it’s growing rapidly. Software robots, basic as they might be, perform well on a certain class of tasks; they can carry out seamlessly structured, information-intensive, easily specifiable, routine work and thus free up employees for more emotive and value-adding activities.
The technology’s main drawback, however, is that it’s not generalizable or transferable. It relies on the explicit encoding of rules and configurations to generate workflows, and, therefore, SW robots can be thrown off track completely by even the slightest modifications in pre-defined process scenarios.
Intelligent process automation (IPA) is an attempt to fix this by injecting some more advanced AI capabilities into RPA solutions. It is a quickly emerging subfield of AI that concerns itself with providing support for automation of long-tail processes that span multiple systems and routines.
In this post, we’ll shed some light on RPA limitations and discuss some of the recent approaches to making the technology smarter.
RPA vs IPA
Legacy applications most huge enterprises still rely on for core operations are rigid, complex and disjointed so, typically, they have an employee responsible for transferring information from one cumbersome application to another manually (by interacting with apps’ graphical user interfaces), which is a tedious and time-consuming task.
RPA, in its conventional form, helps with this to an extent: software bots can record and repeat employees’ interactions with systems’ interfaces and thus remove some high-volume, repetitive tasks from their workloads.
What they can’t do, however, is identify on their own which actions belong to a certain routine (intra-routine learning) or determine which routines are eligible for automation in the first place (inter-routine learning).
The difference between RPA and IPA can be described by the following example. Suppose there’s an IT person who redirects each request (received as a CRM ticket) to an appropriate department. The standard RPA approach to eliminating the task would be to try and create a set of rules for a robot to follow that includes every possible scenario of the routine’s unfolding; but if the process has lots of variants (and requests usually come in various forms), this would quickly become too costly and complex an undertaking.
In IPA, however, we’d try to engineer an ML-powered bot that can identify patterns between requests and departments on its own given a sufficient number of examples (the bot would have to be trained to identify user intent beforehand.)
Teaching the robots to spot highly deterministic tasks that are ripe for automation can, therefore, help companies get rid of tedious work and reduce cycle times; it is also an essential step to making software robots more intelligent.
The post How RPA, AI and IPA Drive Value in Banking appeared first on IT Consulting Company Perfectial.
Top comments (0)