Robots Are Accessing Banking Data Through GUIs Rather Than APIs

Author

Tom Groenfeldt, Contributor

February 1, 2016

Banks and other companies with complex IT systems are turning to robotics to share information without integrating platforms. Instead, they are tapping into data at the user interface level.

Genpact, a global consultancy which is a leader in the field, calls it rapid automation technology, robotics or autonomics.

“It mimics people’s interactions with user interfaces of ERPs, Microsoft Office documents, and databases,” Genpact explains. “Technically speaking, RA interacts with different software systems at the GUI (graphical user interface) or presentation layer, the same level as a human user of the system.”

The advantage is that this form of coupling is lightweight and nonintrusive, so it can be done with code that sits on top of existing systems; it doesn’t requirement digging into existing platforms and integrating them.

“The time and cost involved in typical integration efforts between different software (the most common being workflow/ERP integration) are notorious pain points for operations and IT executives, and RA is a useful additional tool for those issues.”

Unlike human interfaces between systems, robotic software works 24 hours a day, is much faster and not prone to long coffee breaks.

Sanjay Srivastava, chief digital officer at Genpact, said robotics automation is in its early days with huge advances possible as related technologies such as machine learning, semantic computing, natural language processing (NLP), natural language generation (NLG) and other artificial intelligence,  and a design thinking discovery process extend robotic automation deeper into unstructured data.

“Our view is there are about 10 to 12 of these technologies — digital bricks.” They are the building blocks of a solution that can be combined to industrial scale, enterprise grade solutions.

Robotic automation provides results banks need without going in and rewriting heavy-weight applications, he added.

“We don’t see many clients trying to rewrite big clunky system — they want to address regulatory requirements and tap into new market opportunities, like how to go from institutional wealth management to robo advisory. They want to add on capabilities that clients are asking for, but I don’t think they have desire to rewrite core banking or wealth management. The solution has to be agile, something that can overlap current applications, almost a thin film that can sit on top of their existing infrastructure.”

Rather than rewriting a CRM system, a solution would be thin flexible layers sitting above systems of record to manage campaigns or promotions for clients who have multiple accounts that run on different back end systems. These solutions work well in a cloud environment, he added.

“One of the reasons robotic processing operations have taken off is that we don’t get permission to rewrite core apps. The preferred and pure way in finance is API integration, but often this is not possible. Robotic processing cheats — we use the interface layer, not the IT — and emulate what a person would do, but we do it at very high speed to pull the data. It has worked very well because it gets around deep intrusive integration.”

Workflow has been around for ages, but intelligent workflow in an area like insurance claims can determine which claims are simple and should be paid off immediately, which might require review and which should be referred to legal.

“This level of automation through dynamic workflow allows us to optimize the workforce based on competencies and focus areas and speeds up the process of moving claims to the right owners and track claim response time.”

In processing commercial loan applications for clients, Genpact would review structured data like the balance sheet, income statements and other financial information. In addition it will scan the Wall Street Journal and other sources for news like bankruptcy filings, read documents and come to conclusions that it would then write up and put into a recommendation.

“We have now applied technology to deal with the unstructured data, so we can go out and deep crawl through internet sources and use natural language processing to read them and get structured information out of them.”

It’s not easy, he added.

“It’s coming around but it’s still challenging. Once you start pulling in NLP you quickly realize only some makes sense. When looking at unstructured data you get more data than you can use and you have to apply compute intelligence to solve for it by adding qualifiers to tune it down.”

Srivastava thinks he got a glimpse of the future when he missed an Uber car at Heathrow and got charged for it. He emailed a complaint and got a response in two seconds.

“There’s no way a human could have read my entire email and done a response in that time, but it wasn’t a form letter. It was specific to my situation.”

Genpact is looking at ways to use artificial intelligence for retail bank call centers.

“Today we take those calls, look at turnaround time, customer satisfaction, how fast do we answer his question and get him off the line. What is happened in our world now is that the goals are shifting. In some cases you look at how to eliminate the call or avoid a second call.  When you do so many calls you can start predicting the next question so you ask it before the caller gets off the phone.”

Genpact has used artificial intelligence to automate spreading — the exercise of determining risk across a loan portfolio.

“The process has been manual, you bring in analysts and CPAs that go thru the balance sheets and aggregate and do a spreading exercise to come up with risk profile of a portfolio. Now we load up balance sheets — they could be from multiple countries in different languages with different accounting standards — and dynamically aggregate that same risk score with software. It’s much more real-time, we don’t have to wait for a monthly run and when an event happens you can read the impact.”

Looking out two years Srivastava thinks the amount of work people do on computers will change dramatically as artificial intelligence and machine learning become more capable.

“Two years from now if you are calling into a bank, writing into a chat window for a customer service issue or turning in a claim to an insurance company, that will move to a touchless world.”

This article was written by Tom Groenfeldt from Forbes and was legally licensed through the NewsCred publisher network.


Comment this article

Great ! Thanks for your subscription !

You will soon receive the first Content Loop Newsletter