Kasisto's AI-powered smart bot works with more than 20,000 finance institutions to help customers make payments, track spending, and help out if they think they have a gambling problem.
Want to know how much you spent on Uber rides last month? Or at SoulCycle in 2015? A new banking bot called MyKAI can give you some answers.
Kasisto's new banking smart bot, which works with messaging platforms like Facebook Messenger and Slack to make financial transactions and answer questions, is part of a broader trend towards using AI to enrich customers' interactions with companies. You can make payments, ask the bot things like how much money is in your account, and even ask general questions like "What is a CD?" It can even provide an answer if you want to know if you have a gambling problem.
But how does MyKAI fit into the bigger bot world? How is it using AI to enable meaningful conversations with customers? TechRepublic spoke to CEO Zor Gorelov to learn more. Here's the conversation, lightly edited for clarity.
What's the goal behind MyKAI?
We're basically on a mission to help companies to surprise and delight their customers with intelligent conversations. We'd like to go into anywhere in any device, using virtual personal assistants that can be added into mobile applications, or like with MyKAI, on a messaging bot.
The company is a spin-off from SRI International. It's one of the largest research organizations in the world, and AI is the core competency. Siri was one of the latest AI companies that was created at SRI so, in many ways, Kasisto is taking next-generation virtual personal system AI technology to work with them.
SEE: Why Microsoft's 'Tay' AI bot went wrong (TechRepublic)
We started doing conversational systems long before they became fashionable and popularized by recent Facebook and Microsoft announcements. We believe in building systems that combine the conversational AI with deep understanding of a banking domain, because we believe that conversational user interface is becoming mainstream. Clearly, it's a new way for consumers to be connected with products and services.
We think about conversational experience probably a little bit differently than some in the industry. The common thinking in the industry is that there are too many mobile operating systems, too many devices out there, a graphical user interface is hard to do, mobile apps have too many menus and they're hard to navigate.
However, we believe that conversational user interface is about using the power of human language to create completely magical user experiences. We're really trying to redefine how consumers, in this case in MyKAI, interact with banking.
Are customers comfortable sharing banking details with a smart-bot?
What we're learning with MyKAI launch is very interesting. First of all, MyKAI is very secure and very private. We've learned what we needed to learn working with regulators how to protect customers' data, how not to store PII and all that.
With that, what we're learning from MyKAI is consumers are really engaged, and the type of engagement is exactly what we expected them to do. They'd like to understand their spending patterns. We don't know who the users are, we don't know the amount of money in the account—the accounts are all masked out. We take security and privacy very seriously. We tell consumers that we don't store their passwords, we encrypt and anonymize everything, and the engagement that we're seeing is very encouraging.
How does the bot work?
When you sign up for KAI, when you first get notified about KAI, you go to a webpage. On that webpage, you provide your banking information, again this is all pass-through, we don't store any of that information. Say I'm a customer of Chase, here are my online credentials, and in exchange for those credentials we'll give you an anonymized token. That token basically does two things, it anonymously identifies you to a bank. In other words, we don't know who you are as a customer, but once we pass that token, financial institutions are able to map you to real credentials, we never know who the user is. Then the second thing we ask users is to connect if they know they'd like to make payments. The same thing, we get an anonymized token.
The third thing, we allow users to choose how they want to interact with KAI. SMS, Facebook, and Slack are current options. So far, Facebook is the most popular choice. We're seeing many people engaging in Slack, and some are using SMS as well.
What about preventing hackers from getting into the information?
Our access to a KAI user's bank account is read-only. We cannot conduct. This is today, the functionality that we provide in the version of MyKAi for consumers is read-only access to their information.
Can you talk about the technology behind it? Does it use deep learning?
It's a technical system, and it consists of two components. One component is very sophisticated NLP (natural language processing) that is based on a combination of unsupervised and supervised learning. We collect data, we train the system, and the goal of the NLP system is to understand a user's objective. Are you trying to locate a particular transaction? Are you trying to pay somebody? Once we do that, then we have very sophisticated statistical ways of identifying that, based on millions and millions of data points that the system is trained on, specific to banking or finance.
SEE: Machine Learning: The Smart Person's Guide (TechRepublic)
The next step is the reasoning component. The reasoning is very unique because the NLP that they'll use there to help you identify your goal, "I know that's what you're trying to do." The reasoning is really about helping users accomplish that goal. The reasoning is the system that basically says, "Okay, you'd like to make a payment, let me check if you have enough money in your account. Who do you want to pay? Wait a minute, you have two people with that same name in your Venmo/Paypal contacts. Let's see who you want to choose."
The engine enables this conversational experience, and unlike many NLP systems, allows users to complete transactions. Kasisto's MyKAI can get things done as opposed to some of the NLP systems out there that just say, "I understand that you're trying to transfer money. That's great. Go and transfer money." We're really focused on helping complete transactions as well.
Is the end goal simplifying transactions for the user? So they don't have to log into the Chase account, and then into Venmo to pay somebody else, and then into their credit card?
It's not about logging in or not logging in. We're after two things. Somebody recently reported that Americans on average download fewer apps, and more importantly 80% of the time spent daily in apps is spent in three apps, and messaging is one of those. What we're doing by releasing MyKai, we are helping make financial services ubiquitous. In other words, we bring your bank to the app where you spend a lot of your time. Ubiquity is goal number one.
It's what we call lifestyle banking. The idea is, you are planning your life on the messaging platforms by interacting and conversing with your friends, and you're making day-to-day decisions. You're deciding whether it's a good day for you to go out to dinner. Then you need to know how you're doing on your dinner budget this month. By using KAI, you simply go to that contact in your contact list and say, "Hey, KAI, how am I doing on my dinner budget this month?" We say it's a very different paradigm than what existed before KAI—is more you converse with somebody and you make your life plans here. Then you say, "Let me go to my mobile banking app before I make that decision." Log in, and find that balance, or whatever you end up doing.
We're trying to make financial services more ubiquitous and enable lifestyle banking where financial decision-making is embedded in consumers' everyday life.
We've seen a boom in the development of bots lately. How does your stand out?
What I worry about is that a lot of these bots are what I simply call "dumb-bots," which are not incredibly useful to consumers. They're basically glorified key-word searches.
When we launched the banking bot, we picked an area, banking, that is very sensitive, that has a lot of data. We tried to build a smart bot to show to the world what is possible, what kind of conversational experience can be created when the bot that knows about you, knows about your relationship with financial institutions, understands banking. We feel that what we have in the marketplace is quite unique.
SEE: AI experts weigh in on Microsoft CEO's 10 new rules for artificial intelligence (TechRepublic)
I think there is a bot mania going on, there's no question about it. I think the majority of those bots are not incredibly useful. I don't think the world needs another weather bot, for example. I do believe that conversational experiences, whether it's in bots, virtual assistants in apps, or other devices, will continue to grow and become more accepted by the consumers. Conversational interruptions will become more prevalent, and the way we will interact with computing power around us conversationally will become more popular. It does not mean that mobile applications are going away.
With all the respect to Microsoft's CEO, I don't think that bots are a new app. I think apps are good at what they do, and bots or virtual systems will enable new use cases, a new way of interacting between humans and computers.
SEE: Satya Nadella: Software bots will be as big as mobile apps (TechRepublic)
What lessons has your company learned from its journey in developing bots?
We started by building virtual assistants within mobile banking apps, so Siri for banking for lack of a better word. When we started working on bots, and I said this, before bots became popular. Our first thinking was, "Let's just take what we have in mobile apps and move it to messaging platforms. It will be great." It didn't work out that way.
We learned that creating personas and building messaging interactions conversationally into messaging platforms was very different than doing it in mobile applications. We ended up studying behavioral psychology. We ended up hiring people who are writers, the experts in creating dialogue, and we created a persona behind KAI. Not only we taught KAI about banking transactions and payments and all that, we also trained KAI on 1,057 banking terms and we continue to enhance that.
We believe KAI will provide a lot of social and civic utility. It has terrific potential to improve financial literacy of people. It can give people agency over their finances. KAI can respond to questions like, "I have too much debt" or "I have a gambling problem." We feel that this compassion, social, and civic utility in KAI is one of the key attributes. We've invested a lot, and we continue to invest a lot in improving KAI's persona, which is one of the key differentiators for us.
- Slack launches Platform: What Slack users and developers need to know (TechRepublic)
- How to use AI to automatically schedule your appointments with x.ai (TechRepublic)
- Microsoft's Tay AI chatbot goes offline after being taught to be a racist (ZDNet)
- Smart machines are about to run the world: Here's how to prepare (TechRepublic)