Since the advent of online banking services, customers have had several different ways of communicating with their banks. Banks need to monitor all of these incoming customer requests and respond to them in the most efficient way possible. Further, each of the various channels of communication represents a valuable way to segment customers to not only improve how they perceive a bank’s brand but also to market banking products to them better.
That said, customer service teams at large banks may find it difficult to categorize every customer support ticket.
Lee Smallwood, research advisor for our AI in Banking Vendor Scorecard and Capability Map report and COO, Markets and Securities Services North America at Citi, suggests AI may be one solution that banks need to consider when it comes to improving customer experiences:
[When it comes to AI use in banking]…customer experience is what wins in finance regardless of the age or profile of the customer. I think in some aspects of finance, such as micro-loans, customer experience might be millennial-driven, while customer experience in other parts of finance, such as asset management, might not be millennial-driven. But it is certainly customer experience-driven and leveraging AI can allow banks to sift through massive amounts of data to do this.
In this article, we delve deeper into how banks can use AI to improve customer experiences by analyzing the three interviews we conducted this month for our AI in Banking podcast, including discussions about:
- Barriers to AI Adoption for Customer Experience
- The Future of Banking – A Customer’s Perspective
- Chatbots, Conversational Interfaces, and Hype
Special thanks to our four interviewees:
- Yinglian Xie – DataVisor
- Sindhu Joseph – Cognicor
- Sasha Caskey – Kasistobpocas
You can listen to our full playlist of episodes in our “AI for Customer Experience in Banking – Critical Trends and Challenges” playlist from the AI in Banking podcast. This article is based in large part on all three of these interviews:
Subscribe to the AI in Banking podcast wherever you get your podcasts:
We begin our analysis of how banks should go about improving customer experiences with artificial intelligence and what key trends, challenges, and opportunities exist for banks in this space. We start with the top constraints banks face while implementing AI projects for customer service.
Barriers to AI Adoption for Customer Service
Banks might be receiving millions of customer messages every day from the various communication channels they operate. Reading through all of these messages and identifying what issues customers are facing the most is likely far too time-consuming to do effectively.
To this end, banks are now starting to deploy AI solutions that can detect common patterns amongst these messages within them and categorize the data. Banks can then take proactive measures to resolve customer concerns by alerting the relevant department.
Banking leaders need to know that any AI project requires a fairly large-scale data management set-up. Although one might think they’re easier to adopt, AI-based customer service applications can often require more data than a standard AI-based fraud detection system.
Sasha Caskey had some advice for bankers on what type of applications within customer service AI might add value to:
One of the places where AI has been most impactful in banking – taking over mundane repetitive tasks that people have to do. In the form of virtual assistants – AI is shaving off a layer of work that people had to do before.
Making the operations more efficient is what bankers need to focus on. Take something that used to take a long time and reduce it a fraction of the time using AI. Secondly – this enables human customer service reps to focus on areas where they can provide the most value.
Yinglian Xie discussed how both the front- end and back- end customer service processes at banks are not mutually exclusive activities in AI projects:
Let’s take the example of user experience. Businesses want it to be smooth and frictionless. For instance, consumers might be traveling somewhere and have their credit card blocked. This is a type of friction to the consumers. This is the user experience part or the front end, but actually, it’s triggered by the back-end analysis. In the back end, is there enough knowledge derived by analyzing the data to make a correct decision of whether fraud is actually happening or not? This is what determines what happens in the front-end.
Brand Concerns
Customer service AI applications are also challenging because of the constraints that systems need to perform under. For example, an enterprise AI data search and discovery application may return the information an employee is searching for 80% of the time upon initial launch.
Despite the fact that the software will fail to return requested information a fifth of the time, banks may still find the application useful in part because there isn’t much risk associated with the software failing to return that information.
The worst that could happen is employees may need to spend the time searching for it manually, which they were doing before implementing the software. Data scientists can then continue to tweak the software so that it’s even more likely to return the requested information, gradually increasing the software’s efficacy rate to 95% or higher.
A bank, however, may be less likely to want to risk deploying a chatbot that appropriately responds to a customer’s request even 90% of the time. If 10% of their online customers’ interactions with the chatbot are negative, this can reflect poorly on the bank’s brand, which may have lasting effects.
This is especially the case as banking transitions away from in-person relationships with bank employees to wholly digital experiences that may never involve a person-to-person relationship. In other words, banks want to get it right the first time when they introduce new features into the digital customer experience, including artificial intelligence.
Lack of a Culture of Innovation
AI projects for customer service, no matter how big or small, require significant and expensive data management operations. At the same time, all this investment comes with uncertain returns for banks. Bankers have traditionally been highly risk-averse, and a majority of the experts we spoke to agreed that banks need to develop a “culture of innovation”’ throughout their organization, starting with the C-suite. Banks might need to think about what data competencies they can gain at every level in the organization.
Several experts we spoke to for our AI in Banking Vendor Scorecard and Capability Map report repeated this concept of needing to develop an internal data-focused culture at banks. The traditional banking culture might mean that any AI proposals at banks likely need to be “sold” internally to the COO at banks in order to gain approval for funds and resources. Usually, this is a hard sell because of the aforementioned risk-averse nature of bankers. One of the experts we spoke to during our research added:
If a company is looking to work in AI and doesn’t really know where to start, voice of the customer is a great place to start once you’re doing it right because you can get a real impact and you don’t have to explain a lot about what’s going or why things are possible and things like that. You can just get results that allow you to get more AI buy-in throughout an organization.
The Future of Banking – A Customer’s Perspective
Customer Data Applications Might Be Key To AI Dominance
One crucial insight we unearthed from our interviews is that effectively leveraging customer data might be key to larger AI dominance in the future for banks.
This is likely due to the rapidly changing customer preferences in the banking industry and expectations of new millennial customers from their banks. Yinglian Xie gives an example that better illustrates how customer preferences might change and what AI adoption could look like in the future for one application:
Advanced algorithms can play a big role in helping make decisions much faster, which in turn will improve customer experiences. For instance, in identity verification, even if AI helps replace the second factor authentication, the onus of verification is shifted partially to the bank, and customer experiences will be improved. Ideally customers want zero factor authentication, where all the data is safe and yet every customer is easily verified – this is where we might be headed to with AI.
Banks Might Face Stiff Competition from FinTech Firms
Joseph also spoke about how the increase in millennial customers will also affect banks of all sizes in the future. She says these new customers expect their experiences with large banks to be similar to those they have with large tech firms such as Facebook, Google, and Amazon. These internet companies are much better than banks at handling and managing vast volumes of data.
Banks have the advantage of owning a large volume of historical data about banking customers, but banks need to gain data competencies in order to provide similar customer experiences as internet firms.
It might cost a lot more for banks to reach the level of data competencies that the large internet firms have. Sindhu Joseph mentioned what advantage AI is bringing and why banks will need to adapt to survive:
I was at a banking conference – where a use case was presented that claimed – Today’s tech companies can create a huge banking platform at just 10% of the cost that a traditional bank would incur.
Traditional banking needs to become more efficient to service customers they need to increase their digital presence which means streamlining their processes. People don’t need banks anymore – they just need banking. If banks can understand this then they can adapt and survive.
At the same time, banks are now facing competition from smaller FinTech firms, many of which seem to be targeting individual banking services in an attempt to claim some market share from larger banks in that space. FinTech firms and startups have the advantage of having data management policies baked into their business models.
But some of the older banks have been collecting historical customer data for compliance purposes for far longer.
The banking industry involves a lot of compliance regulations and banks of all sizes need to follow strict regulations related to storing and handling customer data. With new regulations, such as GDPR and BASEL IV, it has gotten harder for banks and financial institutions to access and leverage their customer data.
This historical data can potentially help banks understand customer trends that might not be easily available to Fintech firms. But the historical data banks collected wasn’t done with the express purpose of leveraging it to improve customer experiences. Rather, it was stored because banks were required to store it.
As such, this data required a lot of formatting, cleaning, and parsing before it can be fed to machine learning algorithms. Thus, banks will more likely than not need to develop data handling expertise and take a look at how they can move away from siloed data systems to a more AI-conducive data flow.
Big banks with significantly large balance sheets are likely going to invest in FinTech firms, partner with them, buy or acquire them. Smaller banks should look at FinTech firms from a different angle. Fintech firms can be a good indicator of where the future of banking is headed. Smaller banks need to identify these trends and acquire specific data management skills that might be aligned with their business goals.
Chatbots, Conversational Interfaces, and Hype
One of the most widely marketed AI applications in banking is conversational interfaces. Our research shows that banks are marketing their customer service AI applications more than applications for other functions.
That said, AI vendors offering products for customer-facing functions have raised seven times ($1.6 billion to $221 million) less funding than risk-related functions. Thi suggests that banks are not revealing the work they’re doing in the risk-related functions (such as fraud detection) in their press releases; at the same time, it shows how underdeveloped the chatbots these banks are purporting to work on really are.
It is critical for banking leaders to understand what is and isn’t possible with chatbots, virtual assistants and other conversational interfaces.
Despite what some AI vendors might claim, there is currently no out-of-the box chatbot solution that can deliver significant improvements to customer experiences. Chatbots are largely still confined to certain use-cases and answering specific types of questions.
Even a rudimentary level of natural language understanding might require vast volumes of text data for AI systems to learn from. The larger tech firms like Google and Amazon have developed their virtual assistants by using massive volumes of data and even those are still not perfect.
Banks first need to understand what kind of data they have about their customers in order to figure out what their intent might be. Once a chatbot has been deployed, every interaction that it has with a customer is a fresh data point that can be used to tweak and modify the chatbot algorithms to better answer customer questions.
A bank might find that customers from certain geolocations are interacting with the chatbot and facing login issues or issues with ATMs. This can allow banks to proactively take actions to resolve password issues or schedule ATM maintenance.
This is information they can gain through chatbot interactions that might not have been easily possible before.
Victor Pascucci gave another example of how chatbots might add value to banking functions using a customer call center automation use case:
You will see AI embedded in almost everything… lots of work done by people will be better done by AI. For example, the call center agent is probably low paid, is in this position for not much time, and knows little about you as the customer. Think about AI that learns from all the conversations from other customers – knows about you (all your transactions, your social profile)… it would be a super agent… it knows everything about you, about the products that the bank has.
Right now in banking, you have different offerings for different people… you have 3-4 segments of the customer and for each, you have a different offering. If you apply AI on this side of the business you can get very customized offerings for every kind of person.
One additional benefit from using AI and machine learning is that once a solution is deployed, the system can be tweaked to be made more accurate. This can be done using new incoming customer interaction data. This process still would involve the expertise of data scientists to tweak algorithmic models based on fresh data. Chatbots, like most AI applications, will require a lot of maintenance and upkeep to perform effectively.
Chatbots might be able to answer questions up to a certain level of complexity. Sasha Caskey seemed to believe that this is a problem that becomes far more challenging when the goal is to make chatbots capable enough to replace humans:
Virtual assistants are a pretty difficult problem when you get down to it. We might not solve this until the next 20 years. We are able to service a huge amount of customer needs -very expensive to create and maintain, but also pretty effective when you actually create it and put it out there.
The way for banks to ideally approach this challenge of understanding customer intent is to look at what their historical customer interactions have looked like.
For example, banks can use data from surveys or forms filled on websites, incoming sales leads, customer transaction data, and any other sources that might help create a more accurate picture of each customer’s intent. Using AI, banks can develop chatbots to offer services such as account servicing, fetching customer balances or other account details and guiding customers to the information they want on bank websites.
Some banks have also launched chatbots that allow customers to perform transactions by sending messages in the chat window. This is a slightly more complicated feature, and banks might need to think about developing chatbots in such incremental levels of complexity.
This will help them understand customer intent better along the way and in turn, improve the level of understanding that their chatbot algorithms have in discerning customer intent.
Related Articles
- How Chatbots Work, and How They Evolve: In this episode of our AI in Industry podcast, we speak with Yoram Bachrach head of research at Digital Genius about the current applications of chatbots—what they can and can’t do—and details how business can best prepare to automate their customer service
- Artificial Intelligence for UX – Where Bots May Beat Out Brands: In this article, we discuss how AI is being used to develop user experiences. We talk about how AI is becoming a user interface in itself enabling a smarter and more-seamless interaction between humans and machines
- What Chatbots Can Do, and Cannot Do: In this episode of our AI in Industry podcast, we speak with Chief Scientist at Conversica, Dr. Sid J Reddy to discover how chatbots and conversational agents can provide you an advantage in the realms of customer support, product, support, lead engagement, and more, and learn the theory behind creating useful chatbots you can use in your own business
- Banking Chatbots – Comparing 5 Current Applications: In this article we review five banking chatbot applications, comparing their features and offerings.
