5 Fintech Chatbots Transforming the Conversational Banking Experience

Talking finance with a robot is about as complicated as it sounds — unlike the basic live chat experience you might find on a retail website, a good fintech chatbot requires advanced artificial intelligence and machine learning technology for both conversation (natural language processing) and meaningful financial analysis. Driven by these requirements, chatbots in banking have made incredible progress over the last five years, with the best solutions sending personalized, near-lifelike messages that make scripted, rules-based responses look ridiculous. Teams working to introduce or refine their own fintech chatbot technology can learn a lot from the five providers below, whose platforms are transforming the consumer banking experience with each new conversation.

Are you an engineer or product manager looking for the best approach to develop scalable, reliable chat functionality? Try building with the Stream Chat API for free→

Cleo - A Fintech Chatbot for Personal Finance Coaching

The premise behind Cleo — a chatbot that’s not only helpful but wields a snappy personality to match modern internet culture — is aimed squarely at millenials and other highly online users. It’s a fun take on fintech, whose biggest players can still come off as stuffy and old school. Ask Cleo whether tonight’s DoorDash order is in your budget and she’s liable to respond with both a numerical answer and a cheeky comment about your spending habits designed to help you keep your financial goals on track. She’s a prime example of advanced chatbot technology that manages to feel realistic, capable, and fun instead of just spitting back disjointed FAQ answers.

Behind the memes (yes, Cleo will send you memes and even make fun of you if you break your own spending rules), we’re impressed with the level of polish and functionality built into this AI-driven chatbot and its surrounding app. Advanced chat features like emojis and media attachments — not to mention the natural language processing it takes to know when and why a bot should send them — take some serious engineering chops to develop and deliver successfully at scale. The design, chat playbooks, and feature set are all thoughtfully optimized to entertain users, boosting engagement and retention as a result.

Source: meetcleo.com

Unlike other offerings on this list, Cleo isn’t affiliated with a bank and never touches users’ actual money — it just processes the information they provide and returns advice almost instantly. For customers looking to get a better handle on personal finance, it’s a user-friendly mobile alternative to obsessively checking account balances and manually logging purchases. Practical financial planning features include the ability to set and track savings goals, personalize budgets and spending breakdowns, and ask not just whether you can afford a given purchase, but whether it’s a good idea even if you can.

Erica - Bank of America’s Virtual Financial Assistant

Where Cleo makes a point of talking to users like she’s their mildly NSFW best friend, Bank of America’s Erica chatbot comes across with a polite professionalism that still manages to feel thoughtful and human. This bot’s personality may be less memorable, but it more than makes up for that with a highly advanced feature set that boasts deep integrations with Bank of America’s other financial technologies and its users’ bank accounts. According to Tearsheet, Erica was the result of a massive in-house chat build effort, with a team of 100+ engineers and other experts working for 10 months ahead of the 2017 launch and continuing to innovate, refine, and add new features to this day.

Erica isn’t just reading from a pre-programmed script — this chatbot uses machine learning, advanced analytics, and cognitive messaging to adapt based on each conversation and report challenges and unexpected events back to its dev team. With access to Bank of America’s databases, Erica can talk convincingly about a user’s cash flow, upcoming bills, transaction history, balances across multiple accounts, and more. She can catch duplicate charges when they’re posted, confirm that refunds have been credited, remind users to pay bills, and monitor recurring charges — a helpful feature for those of us who could probably stand to cancel an unused streaming service or two. Erica really stands out from the competition by integrating with Zelle payments and applicable Merrill investment accounts, allowing users to access quotes, track performance, and place trades all without leaving the chat conversation.

Eno - Capital One’s Banking Chatbot

Billed as the first natural language chatbot from a U.S. bank, Eno went through an internal beta period to “teach” the AI before launching in 2017. Like other chatbots on this list, Eno uses information from each customer conversation to improve, both on its own and with help from engineers in a supervised machine learning environment. Differentiating features include alerts about suspicious charges and even unusually high tip amounts, a virtual credit card number generator for added security when paying over the internet, and support for SMS messaging, email, and browser-based chat instead of a mobile-app-only experience.

Eno makes for an especially interesting case study in AI and ML chat engineering because Capital One has been more open than other companies in sharing insights their team has learned throughout the development process. From a chat development perspective, a number of steps in Capital One’s process — and lessons their team learned from those steps — stand out.

Capital One VP of Software Engineering Margaret Mayer shared a bit about what it took to teach Eno basic conversation skills and build out the right scalable cloud infrastructure in a blog post around the time of launch. One surprise? Mayer expected that Eno would have to learn around 30-40 different ways customers might ask for their balance. The actual number of unique ways turned out to be 2,200. “To handle this, [Eno’s] natural language processing employs deep learning algorithms and is able to recognize new misspellings and abbreviations that we didn’t previously train it on,” she explains. The team also found that customers were very comfortable using the thumbs-up emoji instead of words to confirm a payment more at least 50% of the time.

Mayer’s other tips for building out reliable, resilient, and scalable chat infrastructure in the cloud will prove valuable to anyone developing in-app chat functionality, whether for a fintech use case or beyond. As we’ve found with chat API customers here at Stream, many developers take their chat infrastructure’s ability to scale for granted and then run into roadblocks as the volume of active users increases. Mayer’s team anticipated this challenge, identifying and fixing an infrastructure problem that occurred during a spike in pilot enrollment: “As we grew, our customers began interacting above our preset threshold...and Eno was not able to respond.” The Capitol One team traced this pain point to the transactions per second limitations at their gateway and has kept that lesson in mind going forward.

In order to support a consistent customer experience across all channels including Capital One’s web and mobile apps in addition to Eno, the team has been investing in an API-based infrastructure. This approach eliminates syncing and data transfer issues, as all front-end customer channels are tied to a common backend API.

KAI - Kasisto’s Digital Banking Engagement Platform

Where other chatbot products on this list are directly customer facing, Kasisto’s KAI platform is unique as a B2B offering that has the potential to help smaller banks and other financial institutions that don’t want to invest the resources to develop their own fintech chatbot from scratch. Kasisto emphasizes KAI’s deep AI reasoning ability, which allows it to parse complex questions and ask relevant followup questions. Many basic chatbots “read” customer messages looking for a single intention to focus on, but KAI can handle realistic conversations that may interweave multiple topics simultaneously.

Kasisto breaks down KAI’s capabilities into six different focus areas, with a separate API to support each. Those six distinct focus areas are Intents, Natural Language, Enterprise User Data, Live Chat Systems, System Usage Data, and Enterprise Management. The chart graphic that shows this structure on the KAI homepage could be a useful example for others looking to architect similar technology and approaches. The core of Kasisto’s platform borrows advanced technology developed by SRI International, the creator of Siri before its acquisition by Apple. Using this technology, KAI can reportedly process over 1,000 potential customer intentions in order to steer a conversation in the right direction.

Businesses using KAI can choose between versions of the platform tailored for consumer banking, business banking, and investment management. The consumer banking product has similar features and functions to the other fintech chatbots in this list, with the ability to check balances, send bill pay reminders, transfer funds between accounts, set goals, break down spending patterns, and provide financial tips. Kasisto offers a number of implementation options — clients can choose between hosted, self-hosted, or hybrid models, with white labeling available.

Credit Karma helps customers monitor and improve their credit scores and compare credit card and loan offers without repetitive hard credit checks, which counterintuitively damage progress. Popular Credit Karma features include a Credit Score Simulator, which helps users predict the effect a given action will have on their credit score without finding out the hard way, and an Approval Odds calculator that processes users’ information to predict the likelihood they’ll be approved for a given credit card (without actually applying). Credit Karma acquired Penny, a popular personal finance app in its own right, in 2018, to add conversational AI to their platform.

Credit Karma’s chatbot is similar to other AI-powered personal finance coaches like Cleo, but it stands out because of Credit Karma’s credit-focused approach (as opposed to simpler personal finance challenges like budgeting). The Penny/Credit Karma integration initially launched with a single feature, in which the bot messaged Credit Karma users to notify them about any new hard inquiries on their credit reports. Then, the team built in functionality to educate users about their credit scores and how to improve them. This fintech chatbot can remember financial goals, answer simple questions, and even offer situational advice based on individual financial information. Users can chat with Penny about credit scores and monthly spending, and also direct disputes right from the chat window.

Building a Custom Fintech Chat Solution

Developing the types of AI and ML-driven chatbot technology used by these five leading solutions requires a serious investment. But that doesn’t mean your team needs to build every component of a fintech chat app from scratch. Whether you’re developing an in-house solution like Erica or Eno or a multi-tenant B2B solution like KAI, the foundation of a great customer experience is a reliable, scalable chat backend that supports the rich feature set today’s users expect.

When you build on top of an enterprise-grade chat API, you get extensible high-performance chat technology and scalable infrastructure that just works. That means your team of fintech experts can stay focused on what they do best: engineering the kinds of financial integrations, technologies, and data insights that set these five stellar solutions apart. Sign up for your free Stream Chat API trial today to see for yourself how easy it is to build chat with the Stream API and SDKs.