Most organizations have either started investing significantly in RPA or are planning to and
according to a Gartner 2019 poll, around 80% of finance leaders have implemented RPA or
are planning to. Unlike digital transformation or cloud migration, RPA is much easier, in
terms of time & money, to adopt. With most of leading RPA platforms investing in cloud
services, Robot-As-A-Service (RaaS) or Robot-for-Rent services will be widely available
and adoption rates will increase significantly.
If you are already using RPA, you are at either step 1
or step 2. This article will focus on the next steps: Intelligent RPA, Self-learning RPA and
Traditionally, an RPA robot takes actions based on
what it is programmed to do, but with the introduction intelligence layer, a Robot can gain the ability to think
and make decisions. There are 3 levels of intelligence:
Configuration Driven: Consists of
rules that tells the robot how to execute at different circumstances. This is usually static data driven
or rules engine driven. This is the most primitive form of intelligence.
Ready-to-Use ML Model: Leveraging
an existing pre-built or pre-configured ML model, most of you would have used these models outside of
RPA implementation. For example, Vision AI by Google and Compute Vision by Azure for image processing or
Google Translate and Microsoft Translator for language translation.
A lot of these generic and popular models
are available on Azure, AWS, Google Cloud, or IBM Watson platform, but industry specific or use case
specific models are also available. So, make sure you spend some time searching before building one.
Custom ML Model: Ready-to-Use ML
model may provide you with a generic model that may be applicable to your non-core business function
(Legal, HR, sales), but most likely, you will have to build a custom model for your core business
functions. The model can be trained to your specific business function and data domains.
You can leverage cloud-based platform
(Azure ML or AWS SageMaker) or an open source platform (TensorFlow or H2O). Most of the leading RPA
platforms have also incorporated ML capabilities to their ecosystem.
An intelligent RPA evolves to a self-learning RPA
when you add the ability of learning. Machine learning works based on the confidence factor, which defines the
probability of an event. Higher confidence means that the result has higher rate to be true positive or false
negative and lower confidence means that the result has higher rate of false positive or true negative.
Through ML model training and testing, you can try
to increase confidence, but once the model is in production and it experiences new data, in some cases, you may
have lower confidence and the robot may fail to execute the operation. In this case, users will step in and
execute the process manually. If you can collect the data on exceptions/failures or successes and you can take
course correction steps by retraining the model, your intelligence will improve over time.
With CD4ML (Continuous Delivery for Machine
Learning) discipline, you can bring continuous delivery principles and practices to the Machine Learning model.
If your organization has adopted DevOps practices, this is not a very complex process.
According to a Gartner report, (Gartner Predicts
2020: RPA Renaissance Driven by Morphing Offerings and Zeal for Operational Excellence) by 2024, organizations
will lower their operational costs by 30% from combining hyperautomation technologies with redesigned
With self-learning RPA, the RPA has now gained
intelligence and can learn, but what is next? The gap is the RPA is executing a process that it was coded for,
and it does not change based on process improvements. In real world, applications get upgraded, new screens or
functionalities are added and business process improves, but the RPA does not automatically upgrade the
With process mining technologies, we can discover
these changes and improve business processes by monitoring system events or user interactions. RPA can use this
information to improvise the automated process or self-heal. RPA vendors, like UiPath with Task mining, has
started providing these capabilities.
Choose 1Rivet for
Your Businesses RPA and AI Needs
1Rivet is a UIPath
partner and we want to explore RPA with you! We are at our best partnering with both business and technology,
collaborating as one unified team. Your success is our top priority! For more information on our focused RPA
solutions, please email us at firstname.lastname@example.org