The Decoding AI™ Podcast
Episode 3: Zor Gorelov, Kasisto CEO, on conversational AI in Finance
Show Notes and Transcript
This episode is sponsored by Teradata.
Teradata is the connected multi-cloud data platform for enterprise analytics company. Our enterprise analytics solve business challenges from start to scale. Only Teradata gives you the flexibility to handle the massive and mixed data workloads of the future, today.
The Teradata Vantage architecture is cloud native, delivered as-a-service, and built on an open ecosystem. These design features make Vantage the ideal platform to optimize price performance in a multi-cloud environment. Learn more at Teradata.com.
In this episode, I sit down with Zor Gorelov, CEO and co-founder of Kasisto, the digital experience platform for financial services, one of the global leaders in conversational AI for finance. A spin-off University of Stanford’s well known SRI, Kasisto has a special pedigree.
Kasisto provides specialist trained virtual assistants to banks and wealth management through their platform, KAI. KAI is trained and ready to host the humanizing experiences that customers seek, while intuitively opening doors to deeper customer engagement. KAI is the industry’s leading Digital Experience platform that was created to master the language of banking and finance. From simple retail transactions to the complex demands of corporate banks and wealth management, KAI transforms simple customer communications into Profound Interactions.
Since my interview with Zor, Kasisto have reported an oversubscribed series C with additional investors bringing in $15.5 Series C extension led in the United States by Fidelity Information Services and internationally by Westpac Banking Corporation, with participation from BankSouth. This brings the total Series C funding to $31 million. And the total funds raised to date at $81.5 million.
Key points discussed in this episode
You really need to know your customer, technology can help.
Accountability of AI - Algorithms have parents
How to approach ethical consideration in conversational AI
Humanizing digital banking experiences
Nuggets of Wisdom
I think the conversation about AI today is for the most part labor augmentation and not labor displacement. We can make great strides in helping advisors work better and be more efficient.
AI can help, but conversational AI is not a silver bullet. You really need to know your customer. Hyper-personalization becomes very, very important.
Again, that goes back to your ”algorithms have parents, right? How the vendor that you work with, how they're looking at the data, analyzing the data and making sure that effects of what is happening in the world are properly reflected in how the assistant is trained.
So deep learning is there, self-training is there, but at the end of the day it's humans who develop those algorithms, humans who analyze data and, and train those systems, and that requires work today, it's not a trivial exercise.
I think the future is in transitioning from convenience driven requests to why and how requests. You can ask your digital assistant, what's my balance? Or what did I spend on Uber? but it's not really that interesting. What is interesting is why is my balance low? How do I improve my balance? Systems can provide unbiased advice in banking and wealth management. So this transition of understanding the user's financial situation, building a 360 degree personalized, hyper personalized view of the user and addressing those types of questions. I think that's what the future is.
Zor Gorelov Linkedin
Links to books and materials mentioned in the episode
How Standard Chartered Uses AI and a Virtual Assistant to Drive Customer Experiences Link
The computer mouse was invented in SRI Link
BBVA’s Lola virtual assistant Link
This transcript has been edited edited for ease of reading and length
[00:00:00] Clara Durodie: Welcome to Decoding AI podcast. I'm here today with Zor Gorelov, the CEO of Kasisto. If I ever need advice on conversational AI in banking, Zor is the person I would turn to. I was introduced to Zor a while back when we met in Singapore, I think it was 2018 or 2019 but definitely before the pandemic. Since meeting, I have followed with great interest and admiration, the work that Kasisto has done to deliver quality conversational AI in banking and in financial services in general.
[00:00:45] Zor, thank you very much for joining me.
[00:00:49] Zor Gorelov: thank you for having me and for the very kind introduction.
[00:00:54] Clara Durodie: in preparation for our interview today, I browsed your website www.kasisto.com. So on the left hand side, as you look at the top of the page, next to the Kasisto logo, there is a word called Enlighten. Tell me about it. I liked it and would like to find out more about it.
[00:01:22] Zor Gorelov: you were lucky to catch our website because, as of Monday, I think next Monday, we're going to launch a new version of the website. So the version you're looking at is about to become an instant classic.
[00:01:37] Clara Durodie: I hope you keep the word Enlighten. I loved it.
[00:01:42] Zor Gorelov: Yes, Enlighten is one of our products and it's a digital assistant named Enlighten because we believe that conversational AI is about surprise and delight. Conversational AI is about using AI to create magical experiences. Experiences that are not available in digital channels. Experiences that are not even available in conversations in the branch or call centers. So we named our product Enlighten because we believe it represents what conversational assistance really needs to do, which is ultimately to surprise and delight users and help them better understand their money, help them make better financial decisions. That's what Enlighten is all about.
[00:02:37] Clara Durodie: The reason for which I love the word Enlighten on a website of a conversational AI company is that it signaled to me that this is a tool you've built and delivered successfully to many large clients, we'll get to that in a minute. This tool manages to humanize the conversational experience. We all need it. When we engage with our banks online, we know how dehumanizing that experience can be in many cases. The other word which stayed with me and piqued my curiosity is when you describe KAI, which is the name of your platform. Your website describes it as KAI gives you the power to transform customer communication into profound interactions.
[00:03:34] I hope you'll keep these two words on your new website because profound interactions are, in my mind, so important when customers call their bank to discuss their money. Why are profound interactions so key? Because we humanize the interaction as we have an emotional relationship with our money. Whether we invest, save, or spend, there is a profound interaction. I think it goes to the heart of some of the core values conversational AI must have in our industry at the moment. Tell me a bit about it, what meaning did you have in mind for “profound interactions”.
[00:04:24] Zor Gorelov: Thank you. Our marketing and sales team thought deeply about what it really means to create profound interactions and why humanizing the digital experience or banking experience is so important. Fundamentally there is, as you say, a digital channel relationship between banks and their customers which is becoming very transactional.
[00:04:46] I'm going to go and check my portfolio. I'm going to go and transfer money, and it's not only transactional, people go to digital properties to check their balance and check their portfolio, especially in today's market because they are worried. So it's not only transactional. It also has some negative connotations, and we believe that using AI, using intelligent digital assistance and conversation, creates an opportunity for banks to help build better relationships with their customers and perhaps re-establish that emotional connection that they need, that is getting closed on. On the flip side, intelligent digital assistance, conversational systems, should help customers make better financial decisions; as you said, financial wellbeing matters.
[00:05:45] I saw some research recently where financial wellbeing was one of the largest contributors to personal wellbeing. That includes health and jobs and family and all that. When people are not doing well financially, all other things tend to suffer. So we feel that there is an opportunity to really change how people interact with their money and interact with financial institutions. Ultimately, what we want is for every user of our systems to use this conversation, to create profound interactions that will help users with their finances over time. We only do financial services, banking, invested management, corporate banking. We don't do any other vertical. We are so focused on this because we believe that domain knowledge and industry knowledge really matters. You cannot have a profound conversation with a general purpose chatbot, it just doesn't happen. So we go deeper and deeper and we're trying to create those experiences that really matter. Since we founded the company, we have had 100 million conversations and every piece of data from those conversations is captured, processed, and labeled. So we are very, very careful about what we do with the data and how we use the data to enhance those conversations. But again, it comes back to those profound interactions.
[00:07:22] Clara Durodie: There was another word which stayed with me which I hope you'll keep in the new website - humanizing. I think this and everything you said about Enlighten and profound interactions underpins the value of delivering these types of engagements. To delight customers and to have profound interactions with them. I think humanizing the engagement is key to delivering for them.
[00:08:19] Clara Durodie: When we try to humanize an agent, a conversational AI agent, how do you humanize a line of code? How do you go about thinking okay I'm gonna humanize my digital agent now. How do you go about that?
[00:08:42] Zor Gorelov: Well, there's a lot that goes into the creation of a digital assistant and creating this human-like experience. We like to say human-like for a variety of reasons. I think that for legal and ethical reasons, people need to know when they’re talking to an AI digital assistant versus a human, but I think some companies are trying to blur the line between the two which we think is wrong. Humans still play a role, but we have a very clear line in the end and say, okay, I am an AI assistant.
[00:09:30] This is what I'm gonna do. I have human-like conversations, but now you have to go to a human and I'm going to hand you over. So we set very clear expectations because there is still a line between the two. Humanizing as per our website is not going to become old. It's going to become a classic. So when the new website goes live on Monday, the old one will become the classic website. We'll keep it in the web archives. Humanizing is definitely one of the components that we're keeping there because it resonates with customers.
[00:10:10] I will tell you a story about two banks. One is a live bank in Emirates that launched Olivia, a digital assistant and another story is about a small bank in the United States, bank south that launched Rita, it's digital assistant. I think the human-like aspect is what matters. Both banks have seen incredible outcomes, incredible results and incredible adoption from their customer base, but also the employees. So I think the human-like aspect is definitely something that we're very focused on.
[00:10:48] Clara Durodie: whilst we are on the topic of humanizing technology and delivering human-like engagement, I quite like the distinction you made i.e. it's a human-like engagement. We don't want to confuse people pretending that the agent interaction is with a human when it's actually not. I quite like my streak on ethical deployment of AI and financial services online nowadays. As a good friend of mine says, you don’t know whether the agent is a fridge, a dog, or a human being, you can be anything you like.
[00:11:34] When talking about our finances, our financial relationship with the world in which we are engaging when we call our bank, I think a human-like experience enables us to have a more comfortable engagement. But at the same time, we just don't want to be fooled into believing that we're talking to a human when we are actually talking to a machine. How do we understand emotions when building an agent like this? Tell me a little bit about how we can design these systems to make them understand human emotions?
[00:12:33] Zor Gorelov: How to truly understand a user's emotions is a very complex topic. You and I are in a podcast. You can hear what I'm saying but you can also see my facial expressions and how I react. Ultimately you can read my reactions which is a lot harder to do in a digital world.
I research knowing language that looking, you know, , looking at profanity, looking at words, users use and how they interact. But I think. Ultimately to truly understand user emotions, and should extend that as possible. You know, they, the year we are looking at each other or we are having a very nice conversation and I may be giving you responses and they may make you angrier, or you may say, well, this is really weird.
[00:13:36] What is she talking about? I don't even understand that. delete audio - this section needs to be deleted from the transcript as well.
So I cannot truly understand your emotion, even when I'm seeing you and this is a lot harder to do in the digital world. I think this is an area that will continue to evolve. It's really easy to say, well, somebody used the four letter word referring to my company. They are upset about something. This is a trivial use case. I think to do it in a more comprehensive way, I think technology will need to continue to evolve. I think user facial expressions will need to be analyzed along with their words. So I think the industry still has a way to go.
[00:14:16] Clara Durodie: Now I'd love to find out more about Kasisto. Since we met in Singapore, I've been following your company's growth. Not only the quality of the systems that you built, but also the company’s leadership.
[00:14:43] As you may know, I spend a lot of time with startup founders and VC investors, trying to find the brightest and the most talented in the FinTech and deep tech space. My submission is that any successful technology is backed by quality management. So the people who decide what kind of tech they will be building are also backed by quality investors supporting that company. From what I read, you successfully closed another funding round for just under 16 million dollars. Tell me a little bit about how you've grown your client base and how you started Kasisto. How it emerged from its academic heritage and what values that brings by implication to our industry.
[00:16:05] Zor Gorelov: we've been around for eight years and venture capital funded. We raised 67 million all together from really top names in FinTech VCs such as Oak Hc/Ft, Two Sigma, Rho Capital as well as strategic investors like MasterCard, DBS, Wells Fargo and other corporations. We have deep roots in the industry and also deep roots into a FinTech investment ecosystem. Very strong investors, a very strong board, and a very strong and experienced management team. Been there, done that. For me, it's this third enterprise software startup with a very strong R&D team that enabled us to attract our co-founder, CTO, Sasha Caskey, formerly with IBM TJ Watson lab. He built a very strong R&D organization around himself.
This is the history of this spinoff from Stanford research. SRI is one of the top five R&D labs in the world. They have a very strong computer science lab, but they're also investing in bioscience, computer science, AI security and robotics. Now why is SRI important?
Despite what Al Gore says, the internet was invented by SRI, the computer mouse was invented by SRI, digital TV was invented by SRI. They have a very strong venture licensing program. One of the first speech companies in the world that was acquired by Microsoft for 20 billion in December of last year was created by SRI.
Apple Siri was created by SRI. And actually, the story goes back to that event where some SRI scientists worked on the next generation virtual assistant platform. So what made it the next generation? Siri is very broad but shallow, right? It knows a lot of things, but it knows very little about every one of those things.
You cannot have a conversation with Siri. I'll give you an example. I like baseball. I follow the New York Yankees. I say, Siri, what was the result of the New York Yankees game? And Siri says Yankees won, Yankees lost. And then I say, what about Boston? And Siri comes back and says it's 57 degrees in Boston.
It's so hard to have a conversation with Siri. This is why the Siri team tries to flip the paradigm rather than the system. So you can ask all kinds of questions. They built a system that could be trained on specific industry language or could be injected with domain knowledge.
[00:19:27] So the system was narrow and specific to a domain, but it was deep. So you could have a conversation. Now, the interesting aspect of that story goes back to the banking industry. The CEO of BBVA at the time, and maybe that's 10 years ago, had a vision of digital banking. He said, look what do people need? How do we engage users on digital, on mobile phones. This is so visionary going back to 2010, 2012 maybe, and the team came back and laid out the vision and there were three key components of that vision. One of those components was human-like interactions that if you want to engage users, you need to create a human-like experience. So, he went to SRI and said, let's build a virtual assistant. Now today, if you Google BBVA space SRI space Lola, you will see that BBVA and SRI built the world's first banking assistant and all of that IP, both SRI IP, and BBVA data and the knowledge became the foundation of this.
So, when we started, we had a very strong foundation for AI on day one. We had all the banking data that we needed. Lola was at the time BBVA's best customer service representative. The vision was to recreate Lola in the AI world, and then BBVA and SRI launched this assistant and they looked at it and they said, this is an amazing creative venture. And that venture ultimately became reality. That's a short history of why we have a strong AI pedigree, a very strong R&D and management team and really amazing investors who provide us with very, very deep connections and an understanding of the industry, both FinTech and banking.
[00:21:47] I think the first time you and I met in Singapore, we were on a panel with Piyush Gupta, the CEO of DBS. DBS was named the best bank in the world a couple of years in a row. McKinsey published a book about the best CEOs in the world and he was named. He is a believer in Ksisto and DBS is an investor as well.
[00:22:20] Clara Durodie: I love how you chose the right words to encapsulate in the heritage, which you actually bring through with KAI, the platform Kasisto has. It's like you build the knowledge and then you bring it the industry, the knowledge and the heritage. Most of the time when we in the industry buy a product, we never really spend too much time thinking about what's behind it. What's the heritage, what's the pedigree. It is also true that there are very few companies like Kasisto, which bring this level of pedigree. It's almost aristocratic pedigree.
[00:23:32] Clara Durodie:
There is another thing which I'd like to dive into today: my background is front office asset and wealth management. That's where I formed my thinking as an industry practitioner, and investment management. Over the years, I haven't seen much fundamental technology progress. But what I've seen KAI doing in the investment management space, it's something which grabbed my attention and made me hopeful that the industry will reach out to you and bring KAI into a work context to deliver what you call hyper personalization. And what I used to call radical personalization.
[00:24:37] Why? Because I believe from my personal experience, as a former practitioner, that you cannot deliver significant progress in advancing digitization in wealth management or investment management if you cannot deliver personalization at scale. So tell me a little bit more, first of all, the areas where KAI delivers technology. I would also like to explore the investment space, because a lot of my followers are from wealth management and asset management. So, they would be very appreciative to find out more.
[00:25:25] Zor Gorelov: Great. So we have our go to market strategy built around digital assistants. We have three in our portfolio. We have a digital assistant for consumer banking, for business banking and for wealth management. They're called KAI consumer banking, KAI business banking and KAI investment management. We started our life in the consumer banking space, then expanded into business banking which covers both small business and treasury services. JP Morgan is our business banking customer. TD Bank and Standard Chartered our customers for our banking assistant. With this experience, we then developed an investment management assistant.
For investment management, we have a customer that uses our investment management assistant to address the needs of both private clients but also advisors. So we are very excited about it. I think it is still early on as to how AI is used, the adoption is not of the scale that we're seeing in consumer banking or business banking.
Forward-looking financial institutions are exploring how conversational AI can deliver financial advice at scale, unbiased and impartial financial advice.
There is also the regulatory aspect to it as well. This is an industry that is heavily regulated and some banks around the world are basically looking to exit that business because they were fined for something somebody did which they couldn't catch in time. I think that conversational AI can be a great equalizer. Conversational AI systems deliver hyper personalized, radically personalized advice which is unbiased. You know, perhaps you start with some idea of optionality. If there is advice like ”I think you should be doing that as opposed to going to buy this product” and somebody calls you back and says, ``I didn't want to buy this product. Why did you sell it to me?” So I think this is an evolving story and I think investment management is a very promising area. Investment management today supports portfolios, positions and equities. Some of the things that you would expect an investment manager would do, but it also supports advisors. So there's somebody asking questions, you call your advisor, correct?
[00:28:21] Clara Durodie: Immediate help when you're on the phone with the client, immediate help is there for you with conversational AI. It's absolutely essential. I can't stress enough how important this help is. When the client phones you, and you're probably at your desk eating, in the middle of your lunch, or drinking a coffee or preparing other work and you literally do not have time to pull out the client file to search for what you need. So, whilst you're on the phone with the client, being able to provide that support in real time is a source of great excitement for me. If I were to work in the front office again, it would be helpful to have that kind of immediate support tool.
[00:29:17] Zor Gorelov: Right. We're doing very interesting work now and in the wealth management space. The problem that exists in wealth management and in fact, many other businesses, is the knowledge and information advisors need exists somewhere. Finding that is not always trivial right. You're on the phone and your customer asks a question about Google shares. So you want to know, hey, what's our recommendation on Google. What's a 52 week high, or who is an analyst covering Google, who I can call. I think this is quite interesting.
They can continue to evolve and become more and more complex. For instance, questions like what is your position on gold? What's the impact on oil prices? There's so much use enriching, perhaps to articulate my view of where AI is today. I think the conversation about AI today is for the most part labor augmentation and not labor displacement. We can make great strides in making those advisors better and more efficient.
[00:30:52] Clara Durodie: Absolutely and I would submit that conversational AI will never replace advisors. Quite frankly, I don't think that could happen. But anyway, that's something for the future to reveal. I'd like to go back a little on the point you mentioned earlier about the value of conversational AI in helping people with financial advice, irrespective of the size of their portfolio. Help everybody irrespective of how much money they have. The reason for which I dwell on this is because I believe there is a very deep untapped business potential to address the mass market, that's the space where typically, at least in the UK, the financial advice doesn't reach. And quite frankly, it might be the people who need the financial advice the most, but in addition to helping such a large, vast majority, it's a new business opportunity which the industry, at least in the UK, haven't been able to tap primarily because of the costs of looking after clients (managing small scale portfolios).
I'd like your help in diving into this space for the benefit of my audience. How can conversational AI unlock business opportunities in a space where historically we haven't been able to reach out due to the cost of maintenance of those customers?
[00:33:00] Zor Gorelov: I think conversational AI and intelligent digital assistance are the only way to really solve that puzzle. The robo-advisors wanted the space, but they were running the same system and using the same algorithm. I think the secret sauce there is hyper or radical personalization. Not only do I need to know that but, I also need to know what assets you have and what are the investable assets. An advisor would need to know what your aspirations are, what you are trying to achieve and try to help hyper-personalize those experiences to better understand users.
When I say understanding, it's not just understanding conversation, but what they're trying to accomplish, having a sort of the 360 degree view of your customers, what are your assets? Where do you, where do you keep them? And I think open banking goes a long way to achieving this, because at least you ask people to have multiple accounts, multiple brokerage accounts.
So understanding the full picture is a really important conversation. AI can help, but conversational AI is not a silver bullet. You really need to know your customer. Hyper-personalization becomes very, very important.
[00:34:14] Clara Durodie: How does Kasisto work in practice? Say I'm one of the many wealth managers based in London. I'm the chairman of the board. Everybody tells me that I need to do something about digitalization of everything and customers expect this. How can you help? How does KAI work and how do I see it implemented in my work that I have been doing for years.
[00:34:57] Zor Gorelov: Say the end state is digital. If you are a digital bank, if your firm has digital apps, mobile apps, it all lives in your mobile app right? So as a wealth manager, you’d have your digital assistant that your clients can interact with. You, the investment manager, would have a digital assistant. In an ideal world, the virtual assistant answers your clients’ questions while you finish your sandwich or finish your coffee. You’d have a dashboard where you can see the conversations digital assistants are having with users, and you are also timely informed of targeted escalations. When somebody's asking my digital assistant an arcane question, one that the digital assistant cannot answer, that'll come into my inbox for me to respond. It's really about the convenience, scale and instantaneous answers. I think many customers, especially young customers, when they have wealth management accounts, don't necessarily want to have a conversation with their wealth manager or relationship manager. They just want to have their question answered. So, how is my portfolio doing, what do I need to know? I think that it's a really great productivity enabler for wealth managers.
[00:36:24] Clara Durodie: Can KAI be tailored to what companies might need? Can they rename it? Not calling it Lola, but maybe, I don't know, they might want another name. Can they make it their own? How does that work?
[00:36:48] Zor Gorelov: Yes. A hundred percent. KAI is powered by a model. It's a branded customizable experience that they can brand and customize but they can also update it to reflect their firm's view of the world and create their own experience. Let's say KAI is a white label end-to-end conversational AI digital assistant platform that firms can customize and make their own like many of our customers have done. I mentioned Emirates, it's a digital bank. Their digital assistant is called Olivia. Standard Chartered Bank, their digital assistant is also powered by KAI and is named Stacy. I think persona design is very important. We didn't speak about it, but part of humanizing your virtual assistant is really about creating a persona that reflects your brand, reflects who you are. I mean, early on, for example, we do a lot of very innovative work around KAI persona, we create multiple personas.
[00:37:52] We decided that KAI is a genderless digital assistant. That was our review. We felt at a time that our customers choose to create male or female personas. Our vision was a genderless assistant, and we did that because we felt that with Siri, even when you change the voice to male, it has feminine personality. We felt that that was wrong, the assistant is not the secretary. So we try to break some stereotypes and create a genderless assistant, and we let our customers customize this persona. So they decide how they want to deploy, but have very specific requirements.
The other thing that we did with KAI early on, was to include social responsibility knowledge into KAI. You'll be surprised. Things that people ask AI assistants are not what they necessarily ask their bankers. We see data in our systems where people say things like I have too much debt. What do I do?. We have people who said that they have gambling problems. So, we looked at the data and we added specific responses and so we are able to direct people to appropriate places for specialist assistance.
[00:39:49] Clara Durodie: The responsibility that comes with the design and deployment of this type of technology hinges on so many conversations. Currently everyone is talking about ESG which is probably here to stay under different nomenclature. Under this narrative, we park the value in identifying biases. Gender bias is one of the biases which has been identified as being impossible to remove.
So, when I hear the creators or the parents of algorithms, like you, are thinking of how you are going to build a piece of tech, a system, which is gendereless, it reinforces my belief that the correction of embedded errors like gender bias is down to the visionary of the leadership, the people behind the tech.
[00:41:01] I coined the line “algorithms have parents” which I believe I said a year or two ago at money 2020 in Amsterdam where we were discussing responsibility on that panel. We are discussing the responsibility system creators have just like we are parents and we influence and shape the thinking and the personality of our children. In the same way, our own ideas, biases, views of life, get transferred into how we design, conceive and implement systems. So, I am very excited to hear about your work on the safety of the systems. I would like to explore this further. You say that you have access to a lot of intimate conversations. People have moments of total honesty, total openness, like I've got too much debt. How am I going to handle that? How can we ensure that the information we gain about people when they engage and they open up to systems ahead of human engagement. How do we ensure that we retain safety? That we protect that intimacy, that level of detailed information about people.
[00:42:39] Zor Gorelov: That's a really excellent question and something that is very important to the industry. If you use an assistant like Siri, or Google assistant, there are all kinds of disclaimers about the data they capture. For instance, they need it for training purposes, but how you, as a consumer, share this data and how the data is used is very, very important.
You also asked questions about biases. I think one of the things that we are doing is operating in a highly regulated and controlled environment. We work as banks, data that we capture gets anonymized, cleansed. We never know who the end users are. So if somebody said, you know how much debt I have? We don't know who that user is. Banks can map that back to their users if they choose to, but we don't know the data. All we know is that I don't know. 20,000 customers around the world ask questions about the financial hardship and then we use it to train the system. Now, as I said, we handled 100 million conversations. We did it in 16 countries around the world in four different languages. We addressed the issue of biases by creating unbiased dataset. I think one of the advantages of using KAI is that it is getting trained, not on a single data set of a single wealth manager with a certain profile of users but it is getting trained on the data across financial institutions around the world. Again, data is anonymized. So we address both of these issues that way; data is really important. We are very careful about the data because it really creates a foundation.
[00:44:45] Clara Durodie: Well, it's the conversation around responsible AI or as I prefer to call it, ethical AI because it's quite black and white and actually puts the conversation into a place where you have to ask, is what I am doing, or not doing as the case may be, ethical. As opposed to, is it responsible or not. This is where the nuances occur when it comes to ethical AI. I think our industry is highly regulated and we know it is learning very quickly to make the distinction. In my view, it is absolutely important to make the distinction between old school compliance, which we're used to having in our industry, and ethical use of technology. I think there are two different ideas, which in some cases, some people and some institutions choose to amalgamate compliance with ethics or compliance with governance and I think they're quite different. However, what is very important, is whilst our industry is still scrambling and trying to work out the difference between old school compliance, which is regulatory compliance, and ethical use of technology, I think it's reassuring for the industry to know that if we buy a piece of tech (Kasisto), they will have ethical issues mitigated and managed to some extent. Ultimately the final result depends on how the systems are maintained once deployed.
[00:46:49] Zor Gorelov: The issue of ethics is quite interesting in how you define it. I feel that ethical AI makes people rethink their own business strategies. I'll give you a very simple example. Today's AI systems can predict certain things. For example, a digital assistant can predict that you're about to miss your payment. Now, as a bank, you can choose to let KAI reach out to the customer and say look, you're about to miss your payment, or you can choose to ignore that and charge that customer with an overdraft fee. So AI ethics questions how banks need to start making decisions about their virtual assistants to handle their customers. And I guess that forward looking banks, banks who want to maintain their customers and grow this relationship will proactively engage with their customers ahead of any potential trouble they face.
[00:48:07] Clara Durodie: It is so true in life too. Help is important when you need it and is available to you.That's when we build trust with the person who offers help. It is the same thing with predicting default. Helping banks to identify when default is imminent and to help the banks help their clients when they're in need. So for instance, it can be KAI saying to the client, which has been identified as an imminent default, saying “just in case you need a payment holiday, we're here to help let us know”. It might be that the conditions are temporarily tight. For that particular person who's been identified as being about to default, to give that person some breathing space when they need it, then that's how we build trust.
In fact, in addition to building trust, I think there's another benefit. There's a monetary benefit to the bank. It's very expensive to deal with clients who have already defaulted. So, why reach that point when it costs you? As a bank, you have to pay someone to go and collect the debt and you lose that client. Just give the client a little bit of help when they need to, they might not even need help after all.
[00:49:47] And we found out through our work that in the majority of default predicted cases, the help is not always needed. There is no need for those people to be helped, but it's good to know. As a customer, my bank reached out to me and said, just in case you need a payment holiday, we are here to help.
[00:50:10] So I'd like to move forward and talk about some of your clients. What do they love about working with Kasisto? Can you share a few ideas? We don't have to go into details but whatever you can share would be much appreciated.
[00:50:38] Zor Gorelov: So we have clients all over the world. They set16 different countries in Africa and IPAC and north America banks of all sizes. We have credit unions that can very cost effectively deploy digital assistance that are on par and or better than banks, Erica. And we have high end customers like JP Morgan, TDBank, Emirates, Manulife, NetBank and Standard Chartered Bank etc.
So really we spend time across the whole spectrum of customers. Our high end customers use our APIs and SDKs and then create their own experiences on a end of the market. We have customers that use no platform to basically configure and go live with a digital assistant very quickly.
I mentioned earlier the one in Emirates, has a digital bank in the UAE has a digital assistant Olivia, powered by KAI. It is a fascinating story, how the bank decided to launch their digital assistant. They started a social media campaign on LinkedIn and Facebook saying that they're about to hire a virtual assistant and it is actually so well done that some people didn't understand that they were looking to deploy an AI assistant. People started applying for jobs. The bank followed it up with a media campaign and naming the contest to create the names of persona design, again, all on social media. Olivia, the name, was born out of personal traits that were assigned to it sort of whole spirit of humanizing a digital assistant. By the time it launched, Olivia had 6,000 people waiting to become friends with her and teach her and train her. All that, and been very, very successful deployed.It's very, very interesting.
Another case study is a small bank in the US in Atlanta. I don't know if they knew about Olivia. We shared some info, but they are a very visionary bank that created a naming campaign. I think they created a female persona. They named her Rita, intelligent teller assistant. It's going to be a very community focused bank. They introduced Rita as one of the digital employees of the bank and said, Rita's here to help all of us. Rita now has its own webpage, its own Twitter account, its own Facebook and interacts with users and banks via mobile phones. And then it automates 96% of all the interactions, which resulted in huge savings. It provides better engagement.
Going back to your question about humanizing. Don't think about launching a chat bot and putting it on my website and I'm gonna move on. I'm going to do something else. Think about creating your next generation of AI powered digital employee, it becomes part of your team, gets trained, gets better and services your customers and the employee digital banker.
[00:54:35] Clara Durodie: Do they see any benefits? Obviously the benefits are clear, but would you be able to share some metrics with us? Just if anything stayed with you.
[00:54:48] Zor Gorelov: I'm happy to, I think that the obvious benefit is reducing servicing costs, right? When we deploy our systems, they usually handle 80% or better of all interactions without requiring human assistance. That results in significant savings in your contact center and your servicing and all that. They also drive better engagement. We have one large customer bank that has 400% increase in digital engagement on their digital and mobile phones. As a result of the deployment of digital assistants, people go in, they ask a question, they get their response. They move on and they do it very, very frequently.
So the results are in the reduction of servicing costs, better digital engagement. And what we're seeing more and more of is we are extending our digital systems, sort of, you know, customer acquisition space, how people open accounts, learn about marketing offers, campaigns, etc. So we're seeing the benefits across all three dimensions, servicing engagement, and acquisition.
[00:56:01] Clara Durodie: How do conversational AI agents or digital agents learn? That's a question I was asked about two weeks ago on a panel. I mean, everybody seems to know how we are going to engage with customers but how do they learn? It's so interesting.
[00:56:23] Zor Gorelov: So, this is an industry that continues to mature, but unfortunately the industry that is full of buzz words and confusion, you talk to people in the space and everything is deep learning, self training, and, you know, it works. It does not really work that way. I will give you a very, very simple and vivid example of that.
You mentioned we met pre-COVID and then COVID happened. Now, we look at interactions of our systems across all our banks, across all countries around the world. The only virus that our systems knew about in February of 2020 was a computer virus. There's no other virus that existed in KAI's knowledge.
But then we look at the data. We started learning that people were asking questions: What is Corona? Is this a COVID virus? So, we started learning and we actually have a slide that shows what we observed in the data. So we started getting the data and retraining KAI.
[00:57:36] They just learned that COVID happened. And all of a sudden, customers started asking us about COVID and the nature of COVID questions changed. They went from What is Corona? to things that you and I talked about earlier to financial hardship like I need to post my mortgage payment because of COVID; or my trip was canceled because of Corona and I need to get my money back.
So those interactions change. And again, that goes back to parents, right? How the vendor that you work with, how they're looking at the data, analyzing the data and making sure that effects of what is happening in the world are properly reflected and assistance in getting trained.
So deep learning is there, self-training is there, but at the end of the day it's humans who develop those algorithms, humans who analyze data and, and train those systems, and that requires work today, it's not a trivial exercise. The reason we raised 67 million is to automate as much of that as possible, but, we still have data scientists and PhDs looking at data, analyzing it and making those assistants better every day.
[00:58:53] Clara Durodie: What is the future of customer interaction?
[00:59:07] Zor Gorelov: I think that a lot of focus in conversation, AI systems over the past several years has been in the area of more efficient data: How's my portfolio doing? Or do I have Apple in my portfolio? What is the cost basis? What did I spend on my last trip to Amsterdam and how much money on Uber? The use cases make it efficient and very convenient to retrieve the data.
I think the future is in transitioning from convenience driven requests to why and how requests. You can ask your digital assistant, what's my balance, but it's not really that interesting. Or what did I spend on Uber? What is interesting is why is my balance low? How do I improve it? How do I improve my portfolio?
So, I think that's the sort of this transition: from what to why and how. Can I afford to go on vacation? Should I be going on vacation? That's where AI systems will really shine. You know, you're not gonna call your bank and say : do you think I should go on vacation?
Systems can provide unbiased advice in banking and wealth management. So this transition of understanding the user's financial situation, building a 360 degree personalized, hyper personalized view of the user and addressing those types of questions. I think that's what the future is.
[01:00:54] Clara Durodie: Talking about the future and buzzwords … everybody is talking about web 3 and metaverse - they are the fashionable things to talk about in 2022. Do you see a space for digital assistants in web 3, How do we engage in that environment?
[01:01:20] Zor Gorelov: I will tell you what we're doing. I'm certainly not a web 3 or crypto expert, but I will tell you what our customers are asking us to do. Many of our customers of all sizes are obviously interested in crypto and that's one of our areas of focus as our customers are making crypto buying and selling available to their own customers. Our mission as a digital assistant is financial literacy. What we are adding to KAI is knowledge about crypto understanding, risks, understanding what crypto is. So when people are going in and like to know what crypto is about, how to execute transactions, what the risks are, prices, they will be able to benefit of our focus in the crypto space.
[01:02:08] Clara Durodie: It seems everybody wants to jump into this space at a very high speed and I've seen some of the smartest people betting quite a lot on Web3. So it's always exciting when everybody's going one in one direction as two things spring to my mind, the first is that I look forward to seeing what they are going to make out of it and the second is I ask Why? because I'm a contrarian. I'm an investor, I'm a contrarian, too. I just wanted to see what's the other way, you know, if I look against the herd movement, what can I see?
Anyways, we are almost an hour into our interview and I'm very, very grateful for your time. Today I've learned a lot. The show notes will include the summary and transcript of the conversation we had today. But before we go…. in my podcast format, I always talk about my guests' background at the end of the podcast. Why? Because I'm a contrarian. Everybody, at the beginning of the podcast, invites their guest to introduce themselves. I get bored right away. I think one's background is important because it's a measure of their life experience and their view of life. But I think that their experience and work is a more accurate measure of what they know. That's why I like to start with that.
So, tell me a little bit about how you managed to reach where you are today with Kasisto. What drove you when you were a 16 year old kid? Did you think that you're gonna run one of the most successful companies in conversational AI?
[01:03:57] Zor Gorelov: no, I didn't think that. I guess you can refer to me as a serial entrepreneur that said this is the third enterprise software startup that I did. I was born and raised in the Soviet Union, in Georgia and went to school there. And in fact, before I became a software engineer by training, I completed a masters in physics and taught myself engineering.
And then I came to the US in 1989 and worked in software companies and then started a software company in the dot com days and continued to do that as co-founding CEO. I love software. I love creating products. I love working with smart people. I love solving problems. My first startup was done in the late eighties.
One day we should get together in New York or London over a glass of wine. And we'll tell you stories about what it was like to run a cash based startup, where we had to fly around the country to buy computer parts and write software.
It was quite interesting, but I sort of got my entrepreneurs and , back in, the day came knowing what to expect on the journey. It's been great. I love what I do and I look forward to more success. I know you, you called us the most successful conversational AI. I think we've done quite well. We have a long way to go.
[01:05:43] Clara Durodie: I tend to have a good hand. When I come to say things about founders and companies so far I haven't missed one instance but next time we are gonna meet, very much look forward to meeting whether it's New York or London, obviously you are based in New York, right? London or New York or whatever around the globe, a signed copy of my book is waiting for you. I have it here on my desk, so it's gonna be in your hands at the first opportunity, Zor. Thank you so much for joining me today. It's been a privilege. It's always a pleasure talking to you and learning from you and I look forward to continuing the conversation.
[01:06:37] Zor Gorelov: It's been a pleasure and let's say learning goes both ways. I always enjoy talking to you and you know, your amazing visibility and network and understanding of the industry. I continue to follow you. And this conversation was really delightful. Thank you for having me.
[01:06:54] Clara Durodie: Thank you and bye for now.
Copying commentaries, articles and transcripts to share with others is a breach of our T&Cs and Copyright Policy. Decoding AI® Newsletter and Podcast Disclaimer.