Robotic process automation and intelligent automation are accelerating, study finds

A new report from Omdia, a global technology research firm, sheds light on how robotic process automation (RPA) and intelligent automation (IA) are gathering increasing momentum in the enterprise technology marketplace.

In the report, “Fundamentals of RPA and Intelligent Automation — 2021,” Omdia defines RPA and IA in the following manner. RPA is where software “bots” mimic humans performing rules-based tasks to improve process efficiencies, quality, and, ultimately, the accuracy of process outcomes. IA incorporates AI technologies such as machine learning and natural language processing (NLP), along with automation solutions such as RPA, to process unstructured data, provide prescriptive analytics, and automate tasks and processes that involve contextual awareness, decision-making, or judgment.

Through an email interview, Cassandra Mooshian, senior analyst for AI and intelligent automation at Omdia, shared her thoughts about a few of the report’s many takeaways.

VentureBeat: Could you explain the difference between RPA and IA, and if there are situations where only RPA will do?

Cassandra Mooshian: The major difference is that RPA is not “intelligent,” so to speak, as it does not include AI. That makes RPA very well-suited for highly repeatable, rules-based tasks that lie within structured data and do not require any decision-making.

Intelligent automation, however, is aimed at tasks that lie within semi-structured and unstructured data that are more variable and require some decision-making, whether that’s able to be done by machine learning, computer vision, and/or NLP or by an employee through a human-in-the-loop or human-over-the-loop type process.

For example, with RPA, one could set a predetermined workflow that sees an email with an invoice attached come through and automatically save that invoice in a (predetermined) specified folder. Intelligent automation can take that further by capturing, categorizing/analyzing, and extracting the data within the invoice such that it can be further processed (this is a common Intelligent Document Processing [IDP] use case).

VentureBeat: Why is it important for organizations to have a strong data architecture before they implement RPA/IA?

Mooshian: It is important for companies to have data management, governance, and security policies in place ahead of implementing RPA/IA as these solutions work across multiple enterprise applications. Ensuring roles-based permissions and access is one component and ensuring data integrity is another, among many others. Automating a broken or error-prone process will not fix the process; rather, it’ll just break faster. For intelligent automation especially, training an ML model on “bad” data can lead to process errors and inefficiencies.

VentureBeat: Low/no-code IA platforms help democratize automation and scale cost and time saving benefits. Could you explain how?

Mooshian: No-code, drag and drop features are increasingly common among IA platforms which are more business user-friendly. Outside of IT and development teams, not many folks know how to write code. But by being able to drag and drop and/or choose from a list, these platforms can be usable/accessible to more users who can create bots or automate a task. The more automation, the more time and cost savings.

VentureBeat: How do process and task mining and intelligent document processing (IDP) extend the reach and effectiveness of process automation solutions?

Mooshian: Process mining is used to obtain a wide lens over business processes and workflows within a company by examining event logs across systems, including how variable they are and where there are bottlenecks. The less variable the process, the greater its potential candidacy for RPA/IA, though other factors must be considered as well.

Task mining is used to understand how a user is interacting with systems and where there are opportunities for automation. Both of the above help identify automation candidates throughout an organization.

IDP is a use case of IA and is growing in popularity, as there are so many document-intensive processes across organizations that impact many employees. IDP has the potential to help save companies a lot of time, and AI models are getting smarter and smarter, further improving IDP outcomes.

VentureBeat: What is the best industry-specific use case of RPA and IA that you have seen?

Mooshian: This is a tough one because there really isn’t one “best,” and a lot of the use cases are horizontal in that they’re applicable to a business function and apply across industries. Top of mind, I’d say financial services has been one of the first-movers regarding uptake of the technology, and there have been many use cases/customer success stories that have and continue to come out of the industry. Some common use cases in financial services are loan processing, mortgage processing, customer onboarding, customer service, and others.

VentureBeat: What are the challenges of weaving RPA and IA into existing IT infrastructures?

Mooshian: Data governance, visibility of shadow deployments (and having guardrails in place for them), and security are all important to set in place ahead of RPA/IA to ensure architectural readiness.

Another challenge is ensuring that the infrastructure is able to handle the increased speed and volume of transactions related to automated processes, whether it’s their own or someone they do business with. For example, many U.S.- based businesses submitted Payroll Protection Program (PPP) loan applications amid the COVID-19 pandemic, some of whom worked with banks that used RPA/IA to do so. While this sped the process for the banks, it ultimately crashed the system of the Small Business Administration, and they were no longer permitted to use this method to submit applications.