top of page
Home: Welcome

The Ro(bot) Says ‘Hello’: conversational AI and chatbots in financial services

Since mid-March, call centers have been overwhelmed. Faced with sudden customer furloughs and layoffs and prodded by the CARES act, financial providers rolled out forbearance and hardship programs for credit cards, auto loans, mortgages, and other services in a matter of several short weeks. Millions are accessing these programs - about 15 million credit-card accounts and 3 million auto loans didn’t get paid in April.

As a result of this overwhelming demand on call centers, banks and finches have been searching for new ways to better serve customers. Conversational AI assistants, which include chatbots, voice assistants, and other bots, represent a new tool. Each is made possible by advances in natural language processing (NLP) and natural language understanding (NLU). These AI-based technologies allow for text and speech data to be analyzed for understanding and to provide a high-quality response back to the user.


In vogue for many years, conversational AI has now firmly entered the financial mainstream. Some financial providers have gone down the path of building their own in-house. Capital One’s Eno (spelled ‘one’ backwards), is one such example. Other providers, meanwhile, have relied on external providers, of which there is an emerging community of providers. Conversational AI and chat chatbot vendors include Ada, Clinc, and Cognizant AI, among many others. Bank of America and its ‘Erica’ voice assistant (short for ‘America’), for example, has relied on a combination of internal and external providers.

Conversational AI is one of several emerging use cases of AI in financial services. To date, AI adoption across key financial use cases has been uneven. This is especially true for use cases like credit scoring that require high levels of scrutiny on account of fairness, bias, or business explainability

Conversational AI is one of the leading use cases for AI in financial services, even as adoption for other use cases, like credit scoring, has been uneven on account of the high level of scrutiny on fairness, bias, and model explainability. Meanwhile, about 70 percent of millennials report positive experiences with chatbots.


I connected with colleague, Kristin Pugliese, who is a DC tech veteran on her experiences deploying an AI chatbot. She joins to share her view on what other financial providers should consider in developing their own.


* * *

Mack: Welcome, Kristin. What do you see as the consumer benefit of conversational AI tools like chatbots?


Kristin: They’re easily accessible and support quick response times. Consumer studies consistently show that these two factors have a huge impact on customer experience. Digital tools like AI chatbots allow a company to give the customers what they want, but remove the challenges that come with staffing 24/7 support teams. This is especially valuable for startups, where teams are often working with limited headcount and neverending ‘To Do’ lists.


Mack: What’s your view of the operational challenge of implementing and managing a conversational AI tool?


Kristin: Content can be tricky. Understanding how your customer base uses your various resources, like FAQs, live support agents, AI chatbot, etc., and how effectively you can direct customers to the most efficient resource is key. From there, you can determine where to surface different types of information. Regardless of the location, it’s important for all content to be cohesive. The individual or team responsible for maintaining the chatbot has to work closely with other support and content teams to ensure a seamless customer experience.


Mack: What advice would you give any fintech company evaluating whether to implement an AI chatbot?


Kristin: Make a data-driven decision! Take a look at how many customer contacts can be solved by simply providing information. If your team members are using their valuable time and skills to simply relay information, your employees and customers will benefit from AI solutions. I recommend doing a cost analysis to compare the cost of implementing AI solutions with the value of the work a team member could be doing in place of answering simple questions.


Mack: What did you find surprising in developing your AI chatbot?


Kristin: Training a bot to adequately communicate with customers around the globe is challenging! It was natural for me to train the bot from my perspective as a native English speaker. However, my customers ranged from students who were still learning English as a second (or third, or fourth…) language to those who were fluent. Not only did I have to draft content that could be easily understood, but I also had to consider how someone who is not comfortable with English may structure their questions. I found it incredibly helpful to study the conversations where my bot failed. It helped me to train different words or phrases to trigger a response than I would have naturally considered.



Subscribe to get this and other Fintalk News

Subscribe Form

Home: Blog Feed
Home: Subscribe
bottom of page