It sounds alluring enough: Intelligent bots step up and do the menial tasks humans would rather (probably) not be doing–or employers would rather they not, in order to slash costs. But streamlining business processes with robotic process automation (RPA) isn’t as easy as setting and forgetting a software bot as if it was an automated rotisserie.
Still, many enterprises forge ahead as if that’s the case. This rush forward is likely one of the top reasons, as consultancy Ernst & Young found, as many as 30% to 50% of initial RPA projects fail.
This isn’t a reflection of RPA technology or its capabilities. It’s generally the result of organizations rushing forward without doing the right upfront work necessary to prepare for heightened software autonomy and automation. To find answers, we reached out to experts to get their thoughts on the most common barriers to the effective enterprise deployment of intelligent automation.
“The number one success factor for intelligent automation is visible sponsorship,” said Jesse Tutt CEO at mattress company Gotta Sleep. Tutt claimed having visible sponsorship is essential because the biggest barrier to RPA success is often resistance from both management and staff.
“Management often have little capacity to allocate time to improvement and secondly resist intelligent automation as they have a perception it may negatively impact staff satisfaction,” said Tutt.
Tutt said the second barrier is the lack of staff with broad skillsets in data, analytics, software development and RPA necessary to succeed. Good data is necessary to support the analytics that will drive the automation.
In addition to data and analytics skills, automation teams also need to be good software developers. Typically, most automation projects are accomplished through direct database or API interaction, Tutt explained.
“Lastly, in cases where systems are antiquated, are only accessible published applications, or which require mouse or keystroke interactions, robotic process automation skills are required,” said Tutt.
Salvatore Stolfo, founder and CTO at Allure Security adds another barrier: the elusiveness of the AI-powered minimum viable product (MVP). Stolfo said, like any complex application, the difficulty is driven predominately by the complexity of the problem, business considerations and computational constraints.
“To achieve an MVP, the performance of the machine learning engine must achieve accuracy levels that make the application work well enough for the intended market solution. There is no magic bullet,” said Stolfo.
Niraj Patel, managing director, artificial intelligence practice at enterprise technology consultancy DMI, said enterprises are floundering when it comes to RPA because they spend insufficient time understanding the customer experience and therefore the ultimate value derived from automation; they lack clean data sources which gets in the way of generating the intelligence; and legacy applications that don’t lend to plug-play of newer modules such as RPA and machine language components; and unrealistic expectations when it comes to robotic software processes.
In part 2 of this story, we will tackle how organizations can beat the barriers to robotic process automation.
