Ready to learn Machine Learning? Browse courses like Robotics Application Machine Learning developed by industry thought leaders and Experfy in Harvard Innovation Lab.
Robotic Process Automation (RPA) is the latest craze among the CIO crowd because, well, robots! Who doesn’t love robots?
The basic idea: get a piece of software to take over the mundane interactions with some existing application. No longer does a human have to click buttons, cut and paste values, or type data into fields.
Instead, the robot will do all that for you! Just think of all the money you’ll save.
Just how smart is that robot? Not to worry – let’s add some artificial intelligence (AI), and presto – now we have Cognitive RPA. Much smarter than that ordinary RPA we had before.
Now our software can make intelligent decisions – judgment calls that heretofore only humans could manage. Forget the paperless office. Now we can have the humanless office, all thanks to robots!
Sound too good to be true? You’re right, at least, for now.
RPA may be getting all the attention, and Cognitive RPA certainly has promise – but the downsides may still outweigh the upsides.
The Pros and Cons of RPA
Deloitte helps set the stage. “RPA software automates repetitive, rules-based processes usually performed by people sitting in front of computers,” explain David Schatsky, managing director and Craig Muraskin, senior managing director – US innovation, both at Deloitte; and Kaushik Iyengar, director – digital transformation and optimization at AT&T, formerly with Deloitte. “By interacting with applications just as a human would, software robots can open email attachments, complete e-forms, record and re-key data, and perform other tasks that mimic human action.”
RPA is especially useful when the interactions are with older, legacy applications. “Technologies like RPA are terrific tools for breathing new life into legacy systems and creating digital process flows, where before there was only spaghetti code, manual workarounds and swamps of data polluting the corporate underbelly,” says Phil Fersht, CEO and Chief Analyst, HFS Research.
Just one problem – if anything changes with the interface, the data, or any other aspect of the legacy app, then the RPA breaks. “Changing interfaces adds complexity to deployment. Because RPA usually interacts with user interfaces, even minor changes to those interfaces may lead to a broken process,” points out Keith L. Murphy, solution architect at OutSystems. “After all, robots can’t adjust their behavior the same way a human would.”
McKinsey has similar concerns. “Changes upstream and downstream, even during bot configuration, can significantly delay bots being put into production,” say Alex Edlich and Vik Sohoni, senior partners at McKinsey. “For example, a new regulation requiring minor changes to an application form could totally throw off months of work in the back office on a bot that’s nearing completion.”
One of RPA’s strengths is also a weakness: the fact that the robots interact with the user interface (UI) of the app. For older apps with no application programming interfaces (APIs), there’s little choice but to interact with the UI – but that approach leads to brittleness.
Most modern applications, in contrast, offer APIs, which give RPA a somewhat more resilient approach to automating interactions with applications. “Today’s RPA vendors are taking a UI approach, not an API approach,” explains Kurt Sand, SVP and GM of the automation business unit at CA Technologies. “If the API is rich enough, however, then that’s the better, more resilient approach.”
CA commissioned a report by Enterprise Management Associates (EMA) to analyze RPA usage. “Those using RPA total 44%, but they are not all enthusiastic about it,” according to EMA’s report Automate or Die – Learn from The Official State of Automation by Dan Twing, president and COO at EMA, which surveyed over 1,000 business and IT professionals at midsize and large firms. “Twenty percent are using bots with current screens, but prefer and are moving toward using more APIs, 14% are redesigning processes specifically to be managed by bots, and 10% are ‘patching up gaps’ in existing processes with bots.”
Even when accessing APIs, RPA may be more trouble than it’s worth. “Many RPA implementations actually fail. Five years into the industry, there are few examples of strong success, among the thousand plus enterprise robot deployments,” says Sanjay Srivastava, chief digital officer at Genpact. “Bots need constant management and maintenance over their productive life.”
AI to the Rescue?
Many RPA vendors are adding AI, or what people are calling ‘cognitive’ capabilities to their offerings. “The integration of cognitive capabilities into robotic process automation platforms has led to the development of Cognitive Robotic Process Automation (CRPA) software bots,” explains BIS Research analyst Rahul Papney. “CRPA platforms can automate perceptual and judgment-based tasks through the integration of multiple cognitive capabilities including, natural language processing, machine learning, and speech recognition.”
Papney goes on to list some of the vendors in this space. “Some of the key companies offering solutions for the RPA/CRPA market are Automation Anywhere, Blue Prism, Nice Systems, Work Fusion, UiPath, Kryon Systems, Softomotive, and Ipsoft, among others.”
Deloitte is sanguine about CRPA. “Enterprises are beginning to employ RPA together with cognitive technologies such as speech recognition, natural language processing, and machine learning to automate perceptual and judgment-based tasks once reserved for humans,” write Schatsky et al. “The integration of cognitive technologies and RPA is extending automation to new areas and can help companies become more efficient and agile as they move down the path of becoming fully digital businesses.”
Yet while promising, CRPA is still in its early days. “The marriage of RPA and AI is still not fully mature,” warns OutSystems’ Murphy. “While the possibility of using RPA with AI to address complex and sophisticated processes is exciting, it’s still in its early stages. This can result in fragmentation and higher setup times.”
The vision of leveraging cognitive technologies to automate business processes is the holy grail for automation generally. For now, however, CRPA doesn’t deal well with real-world scenarios where processes might change. “Then there are security patches and other updates happening discontinuously across all these different versions,” says Genpact’s Srivastava. “Finally, processes also can change weekly; interfacing applications get updated regularly; data formats evolve constantly.”
AI Doesn’t Solve RPA’s Problems
RPA works best when application interfaces are static, processes don’t change, and data formats also remain stable – a combination that is increasingly rare in today’s dynamic, digital environments.
The problems with RPA, however, aren’t that the tools aren’t ‘smart’ enough. Instead, the challenge is more about resilience – dealing with largely unexpected changes in the IT environment.
Adding cognitive capabilities to RPA doesn’t solve these resilience issues – you simply end up with smarter technology that is still just as brittle as before.
The end result: a surprisingly narrow set of use cases where RPA – or even CRPA – can provide substantial business value. “You should consider using RPA if you have a large legacy application as part of a process that functionally works, has no bugs, doesn’t need new features, [and] doesn’t require developing additional applications to support the process,” Murphy warns.
Murphy has one final word of warning. “Being a tactical, simple way to acquire process efficiency gains quickly, RPA may divert attention from strategic and critical projects such as creating new systems to support disruptive business processes or replacing legacy large core systems that are holding you back.”
For our final word, we’ll return to Phil Fersht. “RPA often starts out like a teenage romance,” he quips. “A lot of enthusiastic fumbling around that ends quickly, frequently leading to disappointment.”