The future of robotic process automation isn't what it used to be.
When RPA became mainstream around 2015, cognitive RPA was considered the evolutionary path for the technology. Basic RPA was useful for automating repetitive tasks, but the rules-based approach worked best with structured data. The cognitive variety, however, would add optical character recognition (OCR), natural language processing (NLP) and machine learning to handle semi-structured and unstructured data -- expanding the efficiencies of RPA to a wider range of enterprise activities.
RPA vendors since 2015 have indeed pushed RPA technology in the cognitive direction, adding AI capabilities. But not everyone is convinced this is the ultimate destination for the technology. Ron Schmelzer, managing partner and principal analyst at Cognilytica, a market research and advisory firm that focuses on AI and related fields such as RPA, believes intelligent process automation, or IPA, is the shape of things to come.
IPA diverges from RPA in that it goes far beyond recording a particular business process and repeating it again and again. IPA aims to harness AI to learn how to adjust and improve the process flow, creating intelligent processes.
'Unintelligent at its core'
So, is infusing RPA with NLP and other technologies an exercise in pouring new wine into an old cask? Schmelzer seems to think so.
"Cognitive RPA is still RPA, which means, it's unintelligent at its core," he said. "Simply using OCR or NLP or machine learning approaches doesn't make the systems less cut-and-dried automated as they are now."
Schmelzer isn't alone in this view of the future of robotic process automation. Management consultant McKinsey & Co. last year published an article on IPA, describing it as "an emerging set of new technologies that combines fundamental process redesign with robotic process automation and machine learning."
Intelligent process automation, the article continued, augments the "[t]raditional levers of rule-based automation with decision-making capabilities thanks to advances in deep learning and cognitive technology."
The result is "radically enhanced efficiency," according to McKinsey.
This collection is for CIOs and their organizations researching the business value of RPA, how it's used in IT operations, its current challenges and how it differs from other automation technologies.
The end of RPA?
Smart IPA tools that adapt to new situations and expand beyond templated or recorded activities "will supplant RPA of any sort," Schmeltzer contended.
That possible future of robotic process automation has yet to materialize, however. Cognilytica has created a scale to assess the capabilities of next-generation IPA systems. The yardstick starts at Level 0, which represents RPA products that have yet to embed AI, and peaks at Level 3, which represents technology that can learn from itself to determine better ways to handle a given process flow. At that stage, hard coding disappears and the system relies entirely on machine learning.
According to Cognilytica's assessment, no vendor or technology has reached Level 3.
"We are nowhere near that goal," said Kathleen Walch, managing partner and principal analyst at Cognilytica. "Some vendors have that [high-level IPA] in their vision and are starting to realize the limits of RPA."
Walch said some RPA vendors have reached Level 1 or 2. Level 1 capabilities include NLP and OCR, while Level 2 systems can automatically identify process flows in new systems and mitigate process flow exceptions, according to Cognilytica.
Cognitive RPA as the future of robotic process automation isn't necessarily a dead end for RPA vendor companies. Schmelzer said cognitive RPA tools that "embrace real data science," focus on "how the business process works" and are able to model and discover explicit and implicit business processes, can become intelligent process automation tools.
"IPA does require new technology, but it doesn't require new companies," Schmelzer said.
UiPath, an RPA vendor based in New York, is among the companies building AI into their products. The company has reached Level 2 on Cognilytica's AI scale.
"Neural networking and machine learning … should be embedded as part of any process automation platform," said Param Kahlon, chief product officer at UIpath. "We see that process automation and AI go hand-in-hand."