For the past two years, fed by outsized investments and valuations, a buzz has crackled and fizzed around robotic process automation. It was a bit light and airy, but it sure was hot. Pundits, investors and RPA providers themselves emitted a lot of the heat, leaving some real-world businesses cold.

Last fall, the RPA market leader, UiPath Inc., laid off 11% of its workforce, triggering a wave of speculation: Visible use cases don’t justify the money or the hype; RPA needs more octane from artificial intelligence; and so on. As we begin 2020 — a year when this technology will be pressed to fill its own big shoes — the working world wants to know: Who are you, really, RPA bot?

Though news in recent months has not been 100% flattering to RPA, it hasn’t let too much air out of it either. It’s still the fastest-growing segment of enterprise software in the world, at least according to Gartner Inc.’s most recent tally. RPA revenue grew 63.1% in 2018 to $846 million and will reach $1.3 billion by 2020, according to research Gartner released this past June. This isn’t for nothing. People like the thought of robots — not the physical, R2-D2 kind, but software bots — completing work for them.

Yet many still aren’t clear on what they can accomplish within their companies. They may have seen a vendor’s sky-high valuation or read a blog post on worker replacement. But they don’t know what really happens during a day in the life of an RPA bot.

So what does happen?

The best of times and the worst of times

Proponents tell us that bots will eventually be so ubiquitous, they’ll change the nature of work. They’re bringing about a Fourth Industrial Revolution where rote labor is automated and human brains work all day at peak potential — birthing novel ideas, improving old systems, and inventing new things. Daniel Dines, co-founder and chief executive officer at UiPath, once said there will be a bot for every PC, just as Bill Gates famously predicted there would be a PC on every desktop.

RPA hype segues directly into RPA anxiety. Maybe bots will become able to do some of those brainier tasks. What if, one day, an employee finds that a bot is doing 60% of his or her job? What at first seemed like a gift starts to cast a grim shadow. The company may decide to round up and replace them with the bot. That’s the basic story arc behind fears of bots replacing humans. Yes, there’s the reassuring line about up-leveling people to higher-value work. But it’s been echoing unchanged from RPA evangelists for so long now, isn’t it starting to sound like a line?

In time to stave off a few panic attacks, UiPath laid off roughly 400 employees — about 11% of its workforce — in October. This came months after the startup secured a $568-million series-D funding round that sent its valuation soaring to $7 billion. The widely publicized move was proof to many that the RPA narrative needed redrafting. Was UiPath another early-stage tech company fattened on investor dollars with lots left to prove to customers? Is this a microcosmic event charting the future trajectory of RPA as a whole?

“UiPath is realizing to its cost that intelligent automation is a marathon, not a sprint,” said Phil Fersht, founder and CEO at HFS Research Ltd., as quoted by Information Age. “It pushed the hype around RPA far too aggressively.” 

The facts about UiPath provide ammo for almost anyone with an opinion on it and RPA in general. Yes, its customers grew from 100 in 2016 to more than 5,000 in 2019. It earned $15.7 million in 2017, and annual recurring revenue is projected to reach $300 million by January. True, the company’s chief financial officer, Marie Myers, is exiting amid reports she disapproved of reckless spending. UiPath insists the layoffs don’t indicate distress and are just a timely pruning following rapid growth. Oh, and once, Dines turned down a $1-billion investment from SoftBank Group Corp. leader Masayoshi Son.

There’s a story here with a beginning, a middle, and an edifying ending. Someday, we’ll know it. For now, we might see its outline in what bots are doing in companies right now.

Software in suits: When bots go to work

If we take workers’ word for it, 60% of them spend at least six hours per week — almost a full workday — on repetitive tasks.

“Believe it or not, we have many people who still go through and open up email attachments,” Richard Fong, information-technology manager of finance delivery at Chevron Corp., recently told theCUBE. “That’s their day job. That’s what they do all day long for weeks — usually, maybe about two weeks of doing data processing. They spend the other two weeks doing error corrections.”

Tasks that drain human or other resources with low return are prime candidates for automation. RPA providers like UiPath, Automation Anywhere Inc., and Blue Prism Group PLC offer a range of ways to automate them. There are preprogrammed bots for specific tasks, as well as more flexible, open-ended automation tools. An advantage of preset types is that people can deploy them with little or no coding or developer resources.

Some businesses have shifted low-skill tasks to bots with satisfying outcomes. For example, using UiPath’s platform, Chevron developed a bot that could access email, open attachments, and copy and paste their data into a flat file. Finally, it uploads the file to the enterprise resource planning system.

“That was a big win for us. And that’s just one example,” Fong said.

Automating manual or repetitive tasks could result in a four-day workweek or 60 to 90 minutes more sleep each day for employees. A bot that gets out of bed so workers don’t have to should be warmly welcomed into the company. And coming to work better rested with energy to tackle the hard stuff might, by itself, boost worker productivity.

Many businesses are eager to find out all that RPA can do for them. Fifty-three percent of companies in the 2018 “Deloitte Global RPA Survey” had begun using RPA bots in some form. They achieved payback in under 12 months and an average of 20% full-time equivalent capacity. Beyond the initial plunge, though, only 3% were able to scale beyond 50 bots in production.

Programming and maintenance can be hurdles to adopting and scaling bots. Some think artificial intelligence is the answer — that, with it, bots will somehow take care of themselves. Actually, RPA is a form of limited AI that performs rules-based actions. It’s not the advanced Google-deep-neural-networks kind that appears to think for itself. Will integrating the two produce super bots with greater use-case potential and less need for oversight?

“AI entered the RPA space several years ago as an integral capability, so this is not anything terribly new,” according to James Kobielus, research director and lead analyst at Futurum Research.

RPA bots leverage an AI abstraction layer to automatically infer program code from application elements like screenshots. RPA vendors are now incorporating more advanced AI, like computer vision and sentiment analysis. As the RPA market matures, its AI offerings will grow more sophisticated, according to Kobielus.

“Nevertheless, RPA, here and now, has been delivering considerable results in low-code intelligent process automation and providing powerfully user-friendly task-automation tooling for citizen developers. The case studies reported by RPA vendors and those covered in the trade press amply attest to the technology’s disruptive value,” he said. 

Can brittle RPA fit into agile cloud IT?

IT analyst Jason Bloomberg wrote in Forbes that AI doesn’t address RPA’s main weakness — brittleness. Strict, rules-based automation doesn’t like change. If the user interface, data, or almost any aspect of an app changes, a bot can’t adjust. Cognitive RPA, or CRPA, can automate some judgments with help from natural language processing, machine learning, etc. It may appear smarter within an IT environment for which it was preprogrammed. But even its smarts are brittle and unprepared for changes in the IT environment. And a state of frequent change — think agile cloud, containers, microservices — is where IT is headed.

“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,” said Keith L. Murphy, solution architect at OutSystems Inc.

Banks are often married to legacy systems and apps, which is why they’re thought to be good candidates for RPA. Sixty common business processes make a good fit for RPA, according to this infographic from Auxis LLC. They’re spread throughout organizations from finance to IT and include payroll processing, tax filings, employee onboarding and offboarding, data aggregation and migration, competitor pricing, to name a few.

The price of existing RPA offerings is prohibitive for some companies. There are easy entry points here and there; leading provider Automation Anywhere Inc. offers a free community version, for instance. But free or low-cost, advanced, flexible RPA has generally been lacking. If it were made available to all companies, would they become Petri dishes for creative new use cases?

“The current model of proprietary tools requires businesses to choose from a preset list of possible automation, and our goal is to empower developers to build their own solutions and share those tools with the global developer community and RPA enthusiasts,” said Antti Karjalainen, co-founder and chief executive officer of Robocorp Technologies Inc.

Last month, Robocorp secured a $5.6-million seed investment. The startup, based in San Francisco and Finland, claims to be building the first-ever cloud-native, open-source, license-free platform for RPA development and orchestration. It believes an open ecosystem will do for RPA what GitHub Inc. did for software-development collaboration. 

Proprietary vendors cater to large Fortune 1000 companies, according to Karjalainen. Robocorp wants to make customizable RPA affordable to companies of all sizes. One early user — a bank — automated a single customer-service task with Robocorp’s open-source tools. Within a month, it saved customers nearly 1,400 years of waiting time, Karjalainen pointed out.

“This bank would never have been able to justify implementing traditional RPA tools. But with open-source RPA technology, it was possible,” he added. 

Robocorp sees a future where commonly outsourced work is instead robosourced to bots and developer overseers. This prospect raises a rather large question: How would diverting these jobs from low-wage offshore workers to bots sway the world’s economy?

What of those companies that have already had ample time to exploit RPA? Are their bots saving costs and enabling creative surges day in, day out? How do employees like sharing their desks with bots? At Chevron, employees that used to open attachments have moved onto different roles, according to Fong.

“We have … a cost-avoidance situation. … Because instead of hiring new folks or bringing in contractors, we’re able to redeploy them onto higher value projects,” he said.

Will RPA providers prove it in 2020?

The best place to look for use cases may be inside the offices of automation providers themselves. Who’s had greater access to automation technology over the years? Or engineers more knowledgeable about it? Or more time to test it for different processes?

Automation Anywhere implemented a human-resource bot to produce weekly headcount reports. It saved six hours per week for the employee who’d been generating the reports by hand, according to Max Mancini, executive vice president of the digital worker ecosystem at Automation Anywhere. The company proceeded to build 72 more HR-focused bots.

Mancini has also found a use for RPA at home. He’s building a bot to reconcile his energy credit from Pacific Gas & Electric Co. against the energy report from his solar provider. “There is a lot of data from two different sources, and an Automation Anywhere software bot makes it easy to extract that data, compare and validate it, and then write a summary to an Excel spreadsheet,” he said. 

ServiceNow Inc. does not provide RPA, specifically. However, it does offer related types of automation, like IT processes automation (ITPA), which works with APIs (application program interfaces) rather than user interfaces. The company told theCUBE it has realized gains beyond cost savings through its internal use of automation technology. For example, filing applications for software patents once required management edicts and various time-consuming tasks. After automating those, the resources they consumed shifted to the creative, high-value end of the chain, according to Chris Bedi, chief information officer of ServiceNow.

“We saw an 83% increase in the number of patent applications filed by the engineers,” Bedi said.

At UiPath, interestingly, RPA enabled the hyperactive hiring phase that preceded October’s layoffs. Bots helped it recruit 2,000 employees in 26 countries in two years, according to Guy Kirkwood, the company’s chief evangelist. This is the most impressive thing RPA has achieved within the company, he told theCUBE. Marius Istrate, UiPath’s chief people officer who will exit the company in January, coordinated the project.

When during the last two years did the company realize it might be overstaffed? Given its growth over that time, onboarding some new employees must have been necessary. Even after the layoffs, UiPath still has 50% more employees than it did in January 2019. But it obviously let some go pretty soon after recruiting them. What occasioned the U-turn?

With no dip in market share or revenue, it’s tempting to suggest the layoffs were integral to some larger plan. Do these events trace an arc from mass recruiting for productivity to layoffs to implementation of new projects and processes with bots? A short, intense period of brain harvesting, and then low-cost automation to execute on ideas?

Has UiPath practiced anything like this? And, is it a potential new model for how businesses source human labor? These are “important” questions, according to Kirkwood, who said that UiPath simply had to hire fast to match its growth — with RPA, it got quite good at it. Rapid hiring created organizational inefficiencies, which necessitated the layoffs, he explained.

“Robots deliver productivity gains by making human workers more effective and satisfied,” Kirkwood said. “This will undoubtedly result in shifts of worker roles and positions for every company that uses RPA. Ensuring that no one is left behind, re-skilling and up-[skilling] are therefore critical and a key focus of ours in 2020. But, we don’t see a hiring and trimming cycle as best practice.”

RPA bot: Bringer of a global revolution in work? Or a brittle software tool that agile companies are better off without? Betting it’s the former, providers have updated RPA with some of the same things threatening to outmode it — open source, cloud-native platforming and AI. Will they succeed? It depends largely on whether anyone thinks of anything cool their bots can do at work this year.

Image: Gerd Altmann/Pixabay