Syda Productions - stock.adobe.c
The next phase of robotic process automation is a little hard to pin down.
For some enterprises, the initial wave of projects has stalled, as IT and business leaders look for a better return on investment and reconsider their choice of automation candidates. Other organizations are just getting started down the RPA path and remain largely in the education stage. Vendors in the RPA market, meanwhile, pursue somewhat contradictory tracks: Many are emphasizing ease of use with the goal of placing bot development within the grasp of business users. But some vendors aim to branch into intelligent automation and machine learning, a move that would seem to put RPA back into the hands of technical experts.
Consulting firms, systems integrators and service providers have a varied landscape to navigate as they help clients adopt RPA. Business partners might eventually play a role in helping customers integrate technologies such as machine learning into RPA. But it might take a while before that role becomes commonplace.
Taking a step back
Vadim Tabakman, director of technical evangelism at Nintex, suggested that organizations are regrouping after their initial excitement over RPA. Nintex is a Bellevue, Wash., process management and workflow automation company that works with channel partners.
"I am seeing a little scale back in implementation," Tabakman said, noting that customers, once they have acquired RPA tools, might struggle to determine what they actually need to automate. "That has always been an issue around any kind of automation technology."
He said RPA adopters now recognize the need to slow down, determine their "best bang for the buck" and zero in on where RPA investment makes the most sense.
Finding the best bang for the buck, however, has proven challenging for RPA adopters. RPA deployments, at times, have failed to deliver on customers' ROI expectations. Organizations might hit the proverbial wall when they try to scale RPA deployments.
Greg Betz, senior vice president of data intelligence and automation at NTT Data Services, said the ROI letdown stems from one of two factors: Organizations either targeted the wrong processes or experienced low technology adoption due to lack of training, insufficient executive sponsorship or poor change management.
"The initial wave of RPA projects were sold on the promise of ROI through reduced labor costs that could not be realized, except in cases of high transactional volumes or processes that were very time-consuming in nature," Betz noted.
In addition, RPA marketing has centered on simplicity, which has also contributed to the ROI issue. This ease-of-use positioning resulted in a low-hanging-fruit focus, rather than a strategic, enterprise focus, he added.
Limited, small-scale projects simply haven't added up to an adequate financial return. Craig Le Clair, a vice president at Forrester Research and author of Invisible Robots in the Quiet of the Night, said companies have struggled to create enough "momentum for automation" to pay for their investment in RPA.
Is intelligent RPA the answer?
The difficulty of finding enough of the right automation candidates to justify the cost of RPA has put the spotlight on task and process discovery. Consultants have traditionally taken on the discovery job manually, interviewing users and subject matter experts within an enterprise. But some RPA software vendors have launched automated discovery tools that collect data on how users interact with applications as they go about their daily routines. Machine learning plays a role in discovery, analyzing the collected data and unearthing automation possibilities.
Le Clair uses the phrase "digital worker analytics" to describe the tools that combine machine learning with the ability to record user interactions. He said vendors focusing on attended RPA -- those using the technology to support call center agents, for example -- tend to have stronger digital worker analytics offerings. Enterprises that take the time to use such tools can identify processes ripe for automation in a much higher volume, Le Clair noted.
"The next phase of RPA is using digital worker analytics to discover processes," he said.
In addition to discovering processes, analytics can help organizations standardize them. Process standardization is important, because the initial crop of software bots rejected tasks that stray from standard practice. Le Clair said first-generation RPA didn't handle variation, and as a result, RPA could only automate a limited subset of highly repeatable tasks.
Greg BetzSenior vice president of data intelligence and automation, NTT Data Services
For instance, workers in the corporate finance department might share 20% of the steps of a particular process. So, the bot automating those steps would only address a small piece of the overall process. Analytics, however, might find an opportunity to standardize on 45% of the steps. If workers agree to follow the standard steps in a process, "the bot that we build for that [process] will be much more sustainable," Le Clair said.
Process discovery, or digital worker analytics, is one example where RPA intersects with intelligent technologies. Injecting greater intelligence into the processes, themselves, has been a subject of industry discussion for at least five years. Discussions have taken place around concepts such as cognitive robotic process automation and intelligent process automation.
Intelligent bots are another way to deal with the problem of process variation and further the evolution of the RPA market. Le Clair cited the linking of RPA with more advanced AI components and decision management to handle exceptions. He said the major RPA vendors are "doing the right things" as they decide how much of the newer analytics to embed in their platforms and how much intelligence must be integrated from third-party providers.
Le Clair said RPA consulting and implementation partners will play a critical role in deploying the intelligent components. They might help customers assess the native intelligent capabilities of an RPA platform, introduce their analytics partners to augment the offering and work with clients to "make decisions about what is best of breed," he said.
Tabakman said partners have the technical chops to help customers build bots and make them smarter for improved decision-making and process optimization.
"The partner involved is going to make that part a lot easier," he said. "When it comes to advancing RPA into the AI and [machine learning] space, it is going to be less around the customer doing it themselves, and more around the partners and systems integrators."
Ease of use vs. intelligent RPA
Not every RPA company is buying into the AI future, however. "That is not what we believe is the real opportunity," said Kyle Kim-Hays, chief marketing officer at Softomotive, an RPA vendor.
Kim-Hays, while acknowledging the industry is pursuing AI and greater intelligence, said Softomotive sees the future moving toward attended use-case scenarios. There, the opportunity resides in the "long tail of tasks that will, realistically, never be automated or upgraded because IT departments are always limited in resources."
Softomotive emphasizes ease of use, targeting a "new class of users" who aren't IT professionals, but business users. The company refers to them as "automators." Kim-Hays pointed to the example of KPMG, where tax professionals who attended two 30-minute Softomotive training sessions were able to automate some of their more repetitive and predictable tax preparation tasks.
That type of nontechnical RPA, however, might conflict with increasing use of intelligent technology. "[Machine learning] and AI [are] going to be very specific to highly technical people," unless vendors find a way to make those technologies easy to use, Tabakman said.
Otherwise, adding machine learning will mean writing code, and that would take RPA out of the hands of nontechnical users, he added.
But the RPA market might evolve to accommodate the needs of business users and the push for higher intelligence.
"Ease-of-use functionality should be focused on in RPA products to allow end users to take control of their own productivity for desktop use cases," Betz said. "More complex RPA use cases and, certainly, RPA with AI/[machine learning] capabilities should be left to their IT counterparts."
Not there yet
In any event, intelligence, while making RPA inroads, has yet to sweep through enterprises in a decisive way.
"The integration of RPA with AI/[machine learning] is an ongoing process and one that will mature and expand over the next several years, as there is still a change management cycle to pass through before automation occurs," Betz said. "The evolution will ultimately result in an automation platform that is capable of deriving insight from data and acting on that data immediately. The underlying technologies are still evolving, however, and require a level of adoption and acceptance before they will become commonplace in the enterprise."
One obstacle: Organizations have put AI and automation on parallel tracks. When an AI model suggests a certain action, most enterprises want employees or experts to validate the suggested next-best action, Betz noted. The validation process can take many cycles and involves an adjustment period before AI takes action automatically.
Most companies, Betz said, "still view these as two different ends of the technology journey and have become complacent deploying them as two different software packages or tool sets -- meaning it can take months or years to activate the automation phase."
In addition, companies in some industry sectors are still coming to terms with the basics of RPA, much less intelligent process automation.
Peak Consulting, based in Hickman, Neb., focuses on community banks and the Fiserv core banking system. In that segment of the financial services industry, the challenge isn't adopting AI or rejuvenating stalled automation projects. Community banks are learning about RPA's potential. Peak Consulting partners with Nintex for RPA technology.
Community banks are "still working through what RPA is and how it can be utilized and being able to take the next steps of what it can do," said Cole Anderson, executive vice president and loan consulting director at Peak Consulting. "We find ourselves doing more education on what the product does."
RPA has a foothold among community banks, which typically use the technology for cleaning up and converting data. In that use case, RPA can pull information from different systems and databases and then sift through large volumes of data to root out inconsistencies -- multiple names or addresses representing the same customer, for example. Banks can then feed clean data into their core systems.
Anderson said Peak Consulting's mission is to help banks maximize their investment in core systems, noting that clients might not take full advantage of those systems' features and functionality. The company now plays a similar role with RPA, offering consulting services to help banks understand the scope of the technology's capabilities.
Using RPA to link disparate banking systems is one emerging use case that takes community banks beyond data clean up. Anderson cited the example of loan origination system bridges that connect to core banking systems. Loan origination and core systems come from different vendors. Getting them to talk has required navigating differing middleware systems and data ingestion tools. But RPA can replace those bridges, Anderson said.
"We can inject RPA and … get what we need out of the [loan origination] system and transfer it over," he said.
As for the future, Anderson noted RPA is moving in an AI-driven direction. But his clients aren't ready for that level of adoption. He said banks are nervous when it comes to using AI to make decisions around core bank data. That said, community banks' use of RPA is becoming more comprehensive.
"We are starting to see it penetrate into the market and it is growing," Anderson said.