Robotics: Antidote to “cube farms”?

Within the overarching realm of artificial intelligence, two components work hand in hand: “Machine learning” makes the decisions while “robotic process automation” dutifully carries them out.

That’s a simplified description of where robotic process automation, otherwise known simply as “robotics” or RPA, fits within the AI world.

It’s important to grasp this relationship because many industry observers, including many bankers, are coming to recognize the opportunities RPA presents. These include potentials for cost reductions, increased operational efficiencies, and better customer relationships.

“One of the things that will improve customer experience in terms of speeding things up is using robotic process automation,” says Daniel Latimore, senior vice-president, Celent, in an interview with Banking Exchange. “It reduces cost and mitigates operational risks by reducing errors.”

“Think about those cube farms in any bank of size today,” Latimore continues. “Cubes stretching to the horizon with people in them taking pieces of paper, looking at them, analyzing them, and then doing something where they are given no choice, because it’s prescribed by the rules and processes of what they do. Basically anything that is being done in the cube farm, the RPA folks are trying to turn over to technology.”

Others interviewed express similar views.

“Robotic process automation is at the very edge of the artificial intelligence spectrum, as it has the least amount of intelligence. At the heart of it, it is really rules-based scripting and automation of routine and repeatable tasks,” says Brad Stewart, senior vice-president, head of product, AI Enterprise Solutions, Wells Fargo.

Saying RPA is the least intelligent component under the AI umbrella does not in any way downgrade its importance.

“The [RPA] application is really building tools that automate a special work environment, taking tasks that humans are bogged down with. It’s software tools that automate human tasks in the workplace to achieve consistency,” says Antonis Papatsaras, chief technology officer, SpringCM, an enterprise content management company based in Chicago. Along the way, he says, “You get better quality and increased productivity.”

Michael Abbott, managing director of financial services/digital, North America, for Accenture, describes where RPA fits in with the rest of artificial intelligence.

“First, you have the data,” he says. “Then you interpret the data, and the machine learning is how you systematically learn over time what you should and shouldn’t do. And then the robotics can act on behalf of what you used to have a human do.”

Dealing with docs by the thousands

Speaking about robotics, in this context it has nothing to do with popular images such as Star Wars’ R2D2, Abbott says.

“It’s about, imagine in a mortgage business, you have hundreds of documents coming in. You’re trying to sort them out,” says Abbott. “Think about how [process automation] could do that. First it can understand what’s in those documents. Second you can use artificial intelligence techniques [i.e, machine learning] to comprehend whether it’s a title report or a 1040 … Last but not least you can use robotics to take a look at that document, parse it out, and then push it into different queues or different document holding places in the mortgage process. All of those things in the past were done by people.”

Wipro is a vendor that offers RPA solutions for financial institutions. According to a case study posted on its web page, it worked with a London-based multinational banking and financial services company.

The challenge the bank faced was that it receives more than 1,000 scanned documents each day that need to be indexed. Human agents required 1.5 minutes to process each transaction, picking one document at a time, searching for required fields, and manually keying in the data.

All this was replaced by Wipro’s proprietary systems. Results: 100% accuracy in the transactions processed by robot; 95% of the process itself is automated, the remainder handled as exceptions by humans; the exceptions themselves are automatically routed instead of handled manually.

Not a replacement for IT investment

In December, Celent issued a report in which it studied how RPA is applied in banking operations. One of the report’s conclusions: “RPA is not suitable in all situations, but in cases where a quick, cheap automation solution is appropriate and the underlying systems are not being changed frequently, then RPA may be a useful, pragmatic solution.”

Craig Beattie, Celent analyst and co-author of the report, adds: “RPA can deliver significant business benefits, but it must be applied in the right places. It requires a healthy level of skepticism and pragmatism, and is not a replacement for strategic investment in IT.”

What’s ahead in this area? “Robotics has taken off a lot,” says Sridhar Rajan, robotics and cognitive automation lead for financial services at Deloitte Consultants. “The majority of the large global banks are now well on their way adopting robotics, to improve processes internally by mimicking human tasks.”

Which isn’t to say RPB isn’t scalable to other financial institutions. SpringCM’s Papatsaras points out that there is “a pretty big number of [fintechs and startups] in Silicon Valley that are really working hard with artificial intelligence and providing solutions. So scale is not something that needs to be worried about.

Intriguingly, the December Celent report makes this observation: “Many RPA tools already exist in companies, but they are not labeled as RPA. They are quite easy to deploy. They often don’t require IT involvement, and sometimes they can just be installed on a desktop and left to run. It is worth finding out if these kinds of technologies have been implemented already.”

Which brings up the very likely possibility: There are robots already among us. And they are our friends.

This is the second part of a three-part series about the application of artificial intelligence to banking.

 Part 1: Getting up to speed on AI

 Part 3: What you need to know about machine learning