The Center for Data Innovation spoke to Barney Hussey-Yeo, co-founder and chief executive officer of Cleo, a London-based startup that produces an AI-driven financial advice application. Hussey-Yeo talked about the impact AI can have on how people manage their money, and how making it easier for people to share their data could revolutionize banking.
This interview had been edited for length and clarity.
Nick Wallace: Cleo allows users to manage their finances through a chatbot. Can you tell me a little more about how it works, and what AI makes possible in financial management?
Barney Hussey-Yeo: There are three different components that are really important. The main one is natural language processing, which is being able to take in the unstructured text of somebody’s request and turning that into an intelligent response. That’s really hard, and even five years ago we probably would not have been able to do what we are doing, primarily because it was really hard to get machine learning into production five years ago, whereas now it is easier by several orders of magnitude. There are ways to run models in production and get stuff going, and there are more tools—this is not really fundamental innovation in the algorithms, but it’s everything around that. The natural language processing allows us to take a query from the user and turn it into something intelligent.
Then we have two other components that I think are quite interesting. There’s the transaction data from banks, which is a beautiful source of data, but it is also quite messy. On your bank statement, you get a 40 character string that might say something like “Uber,” but might be some other identifier the user won’t immediately recognize. We turn it all into recognizable merchant names, and then we have another set of classifiers that ask, “well, what the hell is Uber? Why is it important, and what does it do?” So the algorithm will trawl the web and see that Uber is associated with ride sharing, taxis, and cabs—so it’s probably about transport. So we’ve learned a lot there about every merchant across the UK, and many across the U.S.
The third component has to do with how we tell you interesting things about your finances. Are you overspending on your credit card? Is this the correct mortgage for you? All these financial products that people take are generally not optimized in any way, they just take them from the bank they’re with. We offer intelligence there, looking at what you have and looking at what would be better for you, and helping you save money.
Wallace: You used to work as a data scientist at Wonga, the payday loans firm. Wonga has come under a lot of criticism in the UK for its lending practices—did that play a part in your going to work an AI tool that helps people manage their finances? How did Cleo get started?
Hussey-Yeo: Wonga changed my perceptions. I studied machine learning at Bristol. I got my first graduate role as a data scientist at Wonga, and it was there that I learned about financial services, banking—the entire industry. We had so much data and we spent so much time looking at it, at the behavior of customers, at the market, deciding what products to build—and what I found is that banking and financial services are fundamentally broken for a huge swath of the population. Payday lending is a business model that is predicated on people taking multiple loans, which is really bad—but it is such a small problem compared to banking as a whole. Most of the customers coming into Wonga already had an overdraft, they already had multiple credit cards, and eventually their credit scores get so bad that the banks say “no, we’re not going to lend to you any more,” but these guys still need credit! They’re still paying interest, and they’re still spending more than they earn every month. There is nothing that the banks are doing in the market to help you make better decisions or spend less than you earn.
We serve 20-30 year olds, and 60 percent of them have overdrafts and they use credit cards. There’s nothing out there that helps them make better decisions in the long term, nothing that guides them through this sort of stuff. My experience at Wonga left me thinking, “this has to be built, something has to solve this problem.”
But the reality was, I built it for myself: I was earning a decent wage as a data scientist, but I was going into my overdraft every month and thinking to myself, “this is insane.” So I built some little scripts that logged into my online banking and predicted when I would run out of money each month and sent me texts like, “you’re probably going to run out of money, you need to chill out a little bit.” That changed my behavior and the way I thought about money and how I got to payday every month.
That was the kernel of the idea. And then I just built it, really quickly—I went to an incubator called Entrepreneur First, I built the first version in four weeks, got it out there, and just iterated on it. It really resonated with the market: the whole concept of a digital system really resonated with the market. Becoming Cleo wasn’t intentional it was a really natural evolution. I built it for myself, I showed it to my friends who said, “this is useful,” and it took me a few months to even comprehend that it could become a company that could change people’s relationship with money.
Wallace: What kind of insights can tools like this give us into how people spend and manage their finances generally? Have you detected any strange trends or bad habits that could support better financial advice?
Hussey-Yeo: There’s so much. The data you get from a bank is so rich, it’s one of the last bastions of untapped data. Google has search, Facebook has your social data—transaction data is incredibly powerful and there’s a lot you can learn from it. I think there are a few things that are really interesting that we’ve learned.
One is just the human element of how people talk to Cleo as a digital assistant. They ask for advice like they would of a friend. They say please and thank you to Cleo. They want to feel ok about their money, and they ask things like, “is it ok that I spent that much over the weekend, Cleo?” and “how can I save more money?” There are all these very fundamental questions people are asking, that they probably wouldn’t ask even their bank manager or their friends. There’s something about it being an AI and a digital assistant that people trust. I think that’s absolutely fascinating, and it’s a new way of interacting with money.
The other interesting thing about how the vast majority of people’s spending goes to things that are pretty boring. A lot of our customers are in London: the percentage of their income they spend on rent and bills is incredibly high—over 50 percent, which is incredible. There are the common things like Uber and Deliveroo in our top ten merchants, and it is crazy to see the rise of them in the data. But fundamentally, people just want to save more money, and that’s what they ask consistently and constantly in a lot of different formats.
Wallace: The EU’s second Payment Services Directive (PSD2), which goes into effect next year, forces European banks to let customers transfer their data via open APIs. Do you think that will provide a boost to fintech services like yours? For example, is there any data you do not get from banks now that might be useful in financial management?
Hussey-Yeo: PSD2, the new open banking regulation, is going to be an incredible shift. Banks have kept this data locked up where it hasn’t been working for the customer. There is so much you can do to help customers make better decisions, to create better financial products, to increase competition.
It’s a real fundamental shift, because banks have held a lot of power for a long time—there are four big banks in most developed countries. Opening up this data and opening up the APIs creates a chance that the bank need not be the primary interface anymore. It’s an opportunity for startups like us to become the interface for a generation.
If you can have that kind of customer interaction and use that data, you can build something that, on the one hand, serves customers broadly, saves them money, helps them, and increases competition between products and services. But on the other, it’s also the chance to build a really big business where instead of going to your bank’s app, you come to Cleo, and the decisions you make are based on Cleo giving you the best advice. There’s a lot of really interesting companies in this space, but the one that will win is the one that can do the most for the customer, build the most intelligent service, and help people save the most money. And that’s really cool: this is one of the few regulations I’ve seen that is really opening up competition in a market that needs to be more competitive. It’s a great thing, it’s really exciting, and the banks are going to have to fundamentally change their models.
Wallace: Fintech services are becoming more popular, but there’s still a huge distinction between them and traditional banks. How do you think factors like AI and data are going to change banking in the long-term?
Hussey-Yeo: The thing that every fintech wants—and every bank, for that matter—is to become the default place you go to manage your money. Banks aren’t going to die, no one is going to kill banks. Banks will still be there in the background because we need the plumbing they provide, and they’re very good at being regulated. But they’re very bad at the customer experience and serving their customers well. The opportunity of data and AI is to build intelligent services that work for the consumer and save them money. Salaries have stagnated and inflation has outstripped wage growth for a long time, so there’s a fundamental need for people to save money now. There’s a crying need from a generation for something better. The scary thing for banks is the vast majority of their retail banking profits come from lending, overdrafts, credit cards, and getting you in early—while you’re in the playground, almost. They have this concept of customer ownership and lifetime value—that business model is going to have to change if the interface goes to company a like Cleo or Monzo or Revolut.
Individually, we’re never going to offer all of the services a bank does, but we want you to have the best services, so banks are going to work harder. I don’t think banks will ever be the ones leading on that interface, because it is so hard to do that, especially when you’re turning such a big tanker like the banking sector. I think there are opportunities for little startups like us to build the interface and using artificial intelligence to build things that people want. That’s the whole opportunity: build the interface. If banks don’t own that anymore, it’ll be a very different world we live in. The line I give to investors all the time is that if Europe is ever going to build a Google or Facebook sized company, it will be in the next five years and it will be in financial services.