Chatbots are great for personal interactions with banks, but the camera may soon provide an even more powerful tool for gauging customer sentiment. How? By reading the expression on a customer’s face, the way another human would.
Seems unbelievable, right?
Not according to Raghu Rajah, vice president of digital banking, engineering and product management at NCR Corporation. Rajah told Bank Innovation that multiple banks around the world are exploring this technology, known as “computer vision.”
Based in Atlanta, NCR makes automated teller machines, barcode scanners and check processing systems, and also operates a digital banking unit called Digital Insight, which is where the computer vision exploration is taking place. Computer vision, Rajah explained, is a type of deep learning technology employing neural networks. Data enters the network through the device’s camera, and is analyzed. Data is the fuel of AI, and the camera provides data in droves.
“There is a wealth of potential,” Rajah said. “It can be used for mood recognition, or something as specific as identifying if a person prefers tennis to football.”
Computer Vision in the Branch
In the banking world, one area that Rajah is seeing interest for computer vision is retail branch banking. Bank systems can view real-time video of, say, customers on a teller line, and, based on their facial expressions and body movements, get a head start on their pain points or even what service to sell them.
“You would be amazed what a deep learning system can come up with from such an event,” Rajah said.
Speaking of customer pain points, this technology could offer a solution to what Tom McCabe, managing director, global head of transaction banking at DBS Bank, identified as a bank’s wait-and-waste time. To Singapore-based DBS, wait-and-waste time is one of team’s “most important” metric for guiding its innovations efforts, McCabe told Bank Innovation. In a bank branch, wait time refers to the amount of time a customer has to wait until they are served. Waste time refers to the amount of time a customer spends performing repeat actions, like multiple password entries or providing information that the bank already has.
Indeed, computer vision has the potential to take on both these issues.
However, there is one major (and obvious) challenge in deploying computer vision technology: “How do you do this without spooking the customer?” Rajah asked.
NCR is currently working with a number of banks on overcoming this hurdle. While Rajah could not name the banks working on computer vision initiatives, he said that his client roster includes at least one major U.S. bank. However, an important piece of the solution is an established regulatory framework to support it. Until that happens, NCR and others in this space must make do with available AI- and ML-driven tools.
Lucky for them, these areas have plenty of potential.
Artificial Intelligence Beyond Chatbots
Artificial intelligence is using information to predict, or as Rajah says, compute, scenarios. One of the most popular AI applications in banking is chatbots.
“There are banks that want intelligent chatbots,” said Eran Livneh, vice president of marketing for Personetics, which now calls itself the “cognitive banking” company. “But then there are banks like [Royal Bank of Canada] that are taking it to the next level.”
Livneh is referring to RBC’s virtual assistant NOMI, which was developed in partnership with Personetics. With Nomi, RBC got creative by adding a micro-savings feature. So, in addition to providing the customer an analysis of her spending, Nomi directs the customer to save funds in a separate RBC savings account based on their financial profile.
The value here isn’t just about the savings account, Livneh said, but it shows that Nomi can evolve into a bridge between the bank’s retail business and its wealth management business, without having to become a robo-adviser. Nomi can accomplish this simply by directing the customer to the bank’s investment platform as another savings option.
RBC launched the micro-savings capability on Nomi six months ago. Whether it is in the process of using Nomi to connect customers to its investment platform is not known, and Livneh would not offer further details.
RBC did not reply to a request for comment.
Aside from RBC, Personetics is currently working with six of the top 12 banks in North America, Livneh said. It also recently worked with U.K. challenger Metro Bank on its chatbot/PFM application, known as Insights.
Metro Bank’s Digital Director Alex Park told Bank Innovation that aside from Insights, it is exploring AI applications for “certain backend processes at the bank.” For instance, Park said it is looking into AI capabilities in the backend processes of payments where there is much unstructured data that AI and ML can streamline. This will help reduce costs and save employees time for other tasks.
Park, however, acknowledges that banks and fintechs still have a long way to go when it comes to AI.
“There is a lot of talk on AI these days,” he said. “I am a little skeptical about how mature we are as an industry. There is an infrastructure challenge of keeping data integrity, but also we are lacking the regulatory framework.”
There’s a general agreement that more regulation is needed to allow for the full deployment of AI and deep-learning technologies. Yet, IBM’s Watson Financial Services Group is even now using the technology to reduce regulatory compliance costs of banks and FIs in other areas.
AI for Reducing Corporate Banking Costs
IBM’s Watson’s team is developing ways to use cognitive technology to automate and condense a bank’s regulatory compliance and related expenses. Watson General Manager Alistair Rennie told Bank Innovation that banks collectively spend at least $300 billion per year across the world on regulatory compliance.
“A lot of this is done in a manual fashion with no real-time view. This is because most of the systems are legacy and archaic, and produce a lot of false positives.” he said.
False positives are transactions that are flagged by a system as potential frauds or suspicious activity. The legacy systems lack intuitive intelligence, and, therefore, tend to flag more transactions as false positives than a cognitive or self-learning system.
“From a solutions perspective, we’ve narrowed down compliance as a core area for us,” he said.
Other areas Watson is applying AI and ML include fraud detection, privacy, anti-money laundering (AML), know your customer (KYC) and big data. Just last week, IBM acquired data aggregation and analytics company Armanta for an undisclosed amount for this purpose.
But even for IBM, there are certain challenges in the space that are hard to ignore.
“The thing about the AI and ML is you are learning from available data sets,” Rennie said.
This means that AI applications can only be as good as the available data. And in corporate banking especially, “there is a small amount of data in a very complex domain, so there needs to be a system that is able to transfer data in a form that allows such analysis,” Rennie said.
But, overall, artificial intelligence has come a long way since Garry Kasparov battled IBM’s Deep Blue in the mid-1990s.
“In the past four and five years, deep learning has become state of the art,” NCR’s Rajah said. “It will fundamentally change the ways banks solve problems and the way they work get done.”4 - Readers Like This Post