Published in


Banking on Bots

The path to a better customer experience

Conversational AI, including chatbots and voice assistants, are a key component to customer experience within financial institutions.

I had the opportunity to chat with three experts building conversational AI solutions for financial services to learn more about how banks are incorporating these solutions.

The panel included:

Moderator: Arte Merritt, Global Segment Lead Conversational AI, AWS

Watch the video, and read highlights below.

Why conversational AI ?

Many financial institutions started building conversational AI, prior to the Covid19 pandemic, as part of a digital transformation initiative. These initial solutions were high profile, highly personalized virtual assistants — like the Erica chatbot from Bank of America. As the pandemic hit, the need changed as contact centers were under increased pressures. As Cathal McGloin of ServisBOT explains in “how it started, and how it is going,” financial institutions were looking for ways to automate solutions to help get back to “normal” levels of customer service. This resulted in a change from the “future of conversational AI” to a real tactical assistant that can help in customer service.

Haritha Dev of Wells Fargo, saw a similar trend. Banks were originally looking to conversational AI as part of digital transformation to keep up with the times. However, with the pandemic, it has been more about customer retention and customer satisfaction.

In addition, new use cases came about as a result of Covid-19 that accelerated adoption of conversational AI. As Vinita Kumar of Deloitte points out, banks were dealing with an influx of calls about new concerns, like questions around the Paycheck Protection Program (PPP) loans. This resulted in an increase in volume, without enough agents to assist customers, and tipped the scale to incorporate conversational AI.

Selecting use cases

When choosing initial use cases to support, financial institutions often start with high volume, low complexity tasks. For example, password resets, checking account balances, or checking the status of a transaction, as Vinita points out. From there, the use cases can evolve as the banks get more mature in developing conversational AI, and as the customers become more engaged with the solutions.

Cathal indicates another good way for banks to start is looking at use cases that are a pain point, and also do not require a lot of IT support. Some financial institutions may have a multi-year technology roadmap, which can make it harder to get a new service started. A simple chatbot for document collection in an onboarding process can result in high engagement, and a high return on investment. For example, Cathal has a banking customer that implemented a chatbot to capture a driver’s license to be used in the verification process of adding an additional user to an account — it has over 85% engagement with high satisfaction.

An interesting use case Haritha discovered involved educating customers on financial matters. People feel more comfortable asking a chatbot what might be considered a “dumb” question, as the chatbot is less judgmental. Users can be more ambiguous with their questions as well, not knowing the right words to use, as chatbot can help narrow things down.

Another interesting use case involves collections and bill payments. As Vinita indicates, similar to the education use case above, customers may find it easier to tell a chatbot that their payment will be late, or respond to a chatbot asking about a late payment, than to do so with a human. Cathal also sees a lot of interest in chatbots for collections or bill payment issues in the mortgage and forbearance markets as well.

Be were the customers are

The customer experience is a key reason why enterprises are building chatbots and voice assistants. It is not just about saving costs and reducing call-agent volume. It is about being available 24/7 on the channels users prefer to interact. As Haritha states, banks want customers to be able to find the information they need right away.

The solution is not necessarily omni-channel, but being on the right channels. As Vinita explains, the idea is to figure out how to make it so the customer feels like the experience is omnipresent, but to really guide the user to the most efficient and effective means to handle the situation. Chatbots can help with the hand-off and orchestration between modalities.

Multimodal interfaces, i.e. solutions involving both text and voice interfaces, also come into play when there are different use cases that require different forms of engagement. For example, as Cathal points out, if a bank needs a customer to fill out a document, sending a chatbot to capture the document is faster. Multimodal is powerful, not only for the hand-off between automation and live agents, but to capture rich media in a secure way.

With multimodal interfaces, it is important that the contextual information travels. As Haritha explains, there are different levels of support in the journey. When a customer is escalated from a chatbot or Interactive Voice Response (IVR) system, the context about the person and why they are calling needs to pass along to the agent to help provide a more efficient, and personalized experience.

There are regional differences in how customers prefer to interact as well. As Cathal indicates, in Latin America and Asian Pacific, WhatsApp is the dominant channel. Whereas in North America and Europe, users are pushed more towards a mobile app, given it is an environment in which the bank can control the security.

Building trust

Security and trust are important topics when it comes to conversational AI in financial services.

Financial institutions are conservative by nature. Financial fraud is on the rise. Having the chat interface inside a mobile app adds both an additional layer of security, and an additional peace of mind for the user that they are getting the right information from a trusted source, explains Haritha.

Security plays a role in the choice of modality customers select to interact. For example, as Vinita points out, one might be comfortable speaking to a voice assistant by themselves in the car, but as soon as they are in a public setting they may not be comfortable talking about banking. The conversational AI solutions need to be able to handle not only the context and state of the conversation, but the security aspects when customers switch modalities as well.

Part of building trust with the user, is setting clear expectations on what the chatbot can, and cannot do. The chatbot needs to be able to handle solving a high percentage of the issues.

Measuring satisfaction

A common metric financial institutions track in conversational AI is the containment rate. Did the chatbot or IVR resolve the customer’s issue without being escalated to an agent. While this metric is important, one needs to be careful not to optimize for containment at the expense of the user experience. For example, the chatbot could have 100% containment by never escalating anyone, but that would lead to a very poor customer experience.

There needs to be a path to escalation, and chatbots and IVR are a great way to do this as the information collected can be used to better route a customer to the right agent, and help them more efficiently.

Another aspect of containment to keep in mind relates to channel shifting. It is important to monitor if the user comes back on a different channel with the same issue, explains Cathal.

Containment rates can be an indicator of areas of improvement in the conversational design. As Vinita indicates, the containment rate is a precursor to customer satisfaction and can be a “guiding light” to areas of improvement.

There are a variety of ways financial institutions measure customer satisfaction including with CSAT and NPS surveys. In addition, banks may monitor email channels, social media, and perform continuous user testing, as Haritha points out.

Given users are already communicating within the interfaces, conversational AI provides a great way to prompt the user for feedback, directly in the conversation flow. For example, asking the user “was that helpful, or not?”

In addition, one can measure the sentiment of the user, in real time. This can be used to adjust the user path. For example, if the user sentiment is going negative, the chatbot can escalate to a live agent.

Empathy and personalization

As conversational AI continues to mature, one topic that is on the rise is how to handle empathy. How can the experiences be made more personal, or human-like?

As Cathal points out, empathy depends a lot on the language used. This can be more challenging with financial institutions as the language tends to go through lawyers, given their conservativeness and seriousness of the interactions. Cathal is starting to see more flexibility with language used based on treatment paths — not treating every user the same way. The more information known about the user, the more personalized the experience can be as well.

Cathal also makes the point that empathy and personalization are not about passing a Turing Test and being more human, but to provide a more “human-like” service. The chatbot should not try to pretend to be a human, but acknowledge that it is AI, and always give users a path to speak to a human in a few easy steps.

Vinitia echos this point, in Deloitte’s own “bot-sonality” research, it is important to set the expectation that the chatbot is a virtual assistant and not a human.

Haritha also agrees. It should be clear to users they are interacting with a chatbot. Customers are looking for help and support. They want something that is going to give an answer, or a path to an answer. That is the personality financial institutions are striving towards.

Customers are not coming to financial chatbots for entertainment — the conversation is meant to be serious, but approachable. As Haritha describes, the experience is meant to be like an older sister or friend trying to help you, someone on your side.

User sentiment is also a factor in empathy. As Vinita explains, it is important to recognize the user sentiment and tone, and moderate the response. For example a user may be calling in excited because they are cashing a check versus calling in upset or nervous in response to collections.

User demographics related to regional dialects and age can also be a factor in empathy. At Deloitte, conversational designers double as linguists to help guide the Natural Language Processing (NLP) to better understand the dialect nuances that might impede the chatbot from being successful. In addition, as Cathal points out, in some use cases, with older users, there may be a tendency to write, or say, longer messages that cover multiple topics, which can trip up the conversational AI, and result in a bad experience that upsets the user. In these cases, one way to improve the experience is through guided flows, quick replies, and menus, that limit the free form input in some cases to help guide the user.

Future of Conversational AI

Haritha sees conversational AI moving to more personalization in financial service chatbots and voice assistants. She also sees expanding to more channels and connecting the mobile app, website, IVR experiences more together.

Vinita envisions additional technologies from other industries being incorporated into financial services. Interactions will become more realistic, incorporating things like avatars. She also sees new ways to build and fortify trust — customers trusting the AI solution will resolve their issue and trusting the financial institution with confidential information. As the trust goes up, banks will offer more conversational AI solutions.

Cathal sees areas of opportunity for Automatic Speech Recognition (ASR) technology to improve and NLP to handle compound Intents. This will allow for handling more complex requests. At the same time, as financial institutions get comfortable with conversational AI, and see how they really help automate processes, contain calls, and provide a better customer experience, the breadth of what the chatbots can do will grow. To help solve this, it is important to consider multibot architectures to expand functionality in a seamless experience.

As conversational AI technologies continue to mature, it will be interesting to see how financial institutions continue to evolve and implement new solutions.

Arte Merritt leads Conversational AI partner initiatives at AWS. He is a frequent author and speaker on conversational AI and data insights. He was the founder and CEO of the leading analytics platform for conversational AI, leading the company to 20,000 customers, 90B messages processed, and multiple acquisition offers. Arte is an MIT alum.



Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store
Arte Merritt

Arte Merritt


Chatbot, Voice Assistant, and AI Entrepreneur; Conversational AI partnerships at AWS; Former CEO/Co-founder Dashbot