Are Bots Ready to Be Bankers?

During this summer’s AI in Banking Summit we focused on bringing together the industry’s top experts in artificial intelligence and banking. In this interview we spoke with Peter Wannemacher, senior analyst at Forrester Research on his report called “Bots Aren’t Ready to Be Bankers.” Peter breaks down the digital initiatives banks should focus on and creating the best bot down the line. 

Watch our featured interview with Peter Wannemacher.

Listen to the podcast:

Rob: Hi everyone and welcome to the AI in Banking Summit. I’m excited for our next session with Peter Wannemacher. Peter is a senior analyst at Forrester Research specializing in digital strategy for retail banks. He works with executives and banking providers and their partners to help them win, engage, and retain customers. Peter leads Forrester’s ongoing research in digital banking, accessing online digital banking initiatives, and digital sales efforts. Peter, how are you doing sir?

Peter: Pretty well! How are you?

Rob: I’m doing just fine. This session is called, “Are bots ready to be bankers?” Peter, tell us a little bit about yourself and your story.

Peter: Sure! I’m at Forrester now, and I’ve been here for a while. Prior to that, my background was in academia. I sort of came out of a post-grad statistics focus and I came to Forrester as a statistician support role and data scientist. I then fell backward into financial services data and worked with our great group of folks that were analysts, like Cathy Greenberg, Brad Stafford, Bill Royal. I was working with them to help make sure they could support, or in some cases not support their arguments for banking executives with data. About six years ago, I switched roles and I’ve been an analyst ever since.

Rob: That’s great! I’d love to hear more about this. You wrote a Forrester report called “Bots Aren’t Ready To Be Bankers.” Can you elaborate a bit more on the conclusions that you had in that report?

Peter: That report was about a year ago, which in AI time matters a lot. The reason it came about was that a lot of our banking executives, as well as some of their partners, like ad agencies, vendors, traditional vendors, as well as bot vendors, were coming to us and saying, “We need to build a bot.” This what was often happening. It felt a lot like 2007, 2008, and 2009 where banks would come to Forrester and simply say, “We need to build an app,” and we would respond with, “What are you trying to achieve with that app?” Too often the answer was, “We need to build an app.” Yet, I spend a lot of my time doing research in technology and more recently that involves a lot of research into AI. I’ve been lucky enough to work with Google, IBM, and a lot of the other folks and looked into what they are doing in AI. Also, some startups that are doing some really cool things. I am bullish about the long term sophistication and the long term value of AI.

I am bullish about the long term sophistication and the long term value of #AI. @P_Wannemacher #banking. Click To Tweet

People want to build bots and my banking executives want to put money into this and we know that the evolution of AI is rapid, so what’s the answer to this? What should they do and what shouldn’t they do? When we did our research we found a number of factors that were positive, but also a number of factors that were negative. When I say negative, I mean potentially problematic for the bank, for the bank’s customers, and ultimately for the success of both the initiative and the bank’s business outcomes more broadly. I won’t go into all of them, but read the research on https://go.forrester.com/ if you’d like, but if you don’t want to, I’ll just break it down this way. Bot experiences are uneven today, or at least that was true a year ago. At the time, bot experiences across industries, across task, user and journey experiences were uneven. In our research we found that about 70% of the time they were effective and successful, meaning that the customer is able to complete their objectives and do so conveniently. 70% is a great percentage and for firms like H&M and Taco Bell. It’s also a useful way of using technology to improve customer experiences, while reducing costs per interaction.

We found that 70% of the time #bots were effective and successful. @P_Wannemacher #banking. Click To Tweet

The problem is that 30% is really high for banks and add to that the fact that people’s money is an especially sensitive subject. So what we intended to say is that if you’re using Taco Bell’s TacoBot on Slack, which is a great bot, if the conversation with the bot goes a little flimsy and you end ordering six extra tacos, that might be kind of annoying and you may have to pay an extra couple of bucks, but it’s tacos. In contrast, if you have no way of understanding how much money you have in your account, or if you accidentally move money multiple times when you only meant to move it once. This could be because of a broken bot, a misunderstood bot or human. The stakes for this can be really high in reality, or even if you put aside the real consequences, the perceived consequences from the standpoint of the customer can be really high.

I’ll quickly end this part of this by saying this got a lot of press, some good and some bad. I did get a lot of good press, but then a lot of bad reaction, which I’m happy to talk about. One journalist had their headline and I think it’s better than our writing. There headline was this, “Financial Services Aren’t Tacos.” I thought, ‘that’s exactly right’ and that’s what it comes down to. No matter how big the bank is they do not have unlimited resources and money to spend. So when it came to what should you invest your money into today, the answer wasn’t nothing that has anything to do with bots. The answer for most banks was that you don’t want to use your money trying to rollout a general purpose bot that will give your customers uneven experiences. Instead, you can use that money to invest in those backend, foundational, and digital initiatives that will enable you to have the best bot in the world a year from now, two years or even three years from now. This horizon differs by bank, but that was the thesis.

Invest money in digital initiatives that’ll enable you to have the best #bot down the line. #banking Click To Tweet

Rob: It’s interesting because we took a look at the report and tried to distill some of the information from that. As an individual building a business based on artificial intelligence and creating these kind of assistances for banks, I kind of took a step back and said, “How do you build data assets within a bank that can help them in these initiatives, if you’re not putting effort into the technology?” I think you’re absolutely right. How do you build a foundational data bank and data layer that anything can start to connect. It doesn’t matter whether it’s a bot, app or website, but instead how it is that you lower the time, effort and innovation cycles it takes to push these technologies out. I really find it interesting with where you go with it. To be honest with you, when we work with banks the biggest initiative that they are always under is the consolidation of that data layer, trying to understand how to maintain governments and access to this data in a clear and concise way. I agree with you concept of taking a step back, building a foundation and then start to increase the rapid pace of innovation. AI as an initiative is about building data assets and the ability to create those data assets. If you don’t do that now it can set you back in the future, but it doesn’t mean that you should ignore the fundamental issues that every bank deals with, which is the data layer. I love that concept that you guys were tying into.

In order to build data assets in a bank you need to put effort into the technology. #banking Click To Tweet

Peter: Going back to the analogy of apps when they first arrived with the iPhone first launched, it was a big deal. Ultimately a lot of the banks built apps and our objection was never t to say, “Don’t ever build that app.” The question was do so properly, purposefully and make sure that you have the backend that can make it a great experience for customers. Going back to the bots question, I’ve actually worked with many banks since that report where we’ve helped them build a bot. There’s truly two questions, what does the bot do and who does it do it with and for. I still worry about this idea of a bot as teller, which we may get to, but this idea that the bot is a generalist for all your banking needs even the firms that have done it pretty well like DBS, USAA, and Ally. Even for those banks, it’s still uneven. If you’re going to be half as good as USAA’s bot then you are not going to be good enough. Instead we urge banks to start building bots that are focused on specific needs and purposes. You’re building out those data sets, which are the most important aspect of it, and governance.

Banks need to build #bots that are focused on specific needs and purposes. @P_Wannemacher #banking #fintech Click To Tweet

Right now, no bank is fully lacking, but there is a maturity level that’s too low and so we see those leading firms that are ready to roll out a bot, either as a focused or generalized bot are usually well ahead on the maturity curve. What we’ve found is that a lot of them are trying to build these bots with specific needs and purposes. An example I like is debt. A lot of people have debt, as we hear about in the news all the time, especially young people who are often digitally savvy and are willing to experiment with a new UI or whatever it may be. Humans, of all ages, don’t like talking about debt with other humanes. Even humans they don’t know. Talking to Barbara in Kansas City is something that causes anxiety and you can actually see in the bank’s data that people try to avoid it. A bot might be a really great fix for that because the anonymity of the bot, the non-humanness of the bot is valuable and not all debt questions are straightforward, but debt is a very specific part of your financial life. A good bot, with great “brains” behind it and data assets is something I’m really excited to see. I’m really excited to start writing about bots that are doing really well. All these concerns about needing a bot with press around it still worries me, but I’m excited to see this evolve and is something I’m very bullish on.

A #bot with great “brains” and data assets is something I’m excited to see. @P_Wannemacher #banking #fintech Click To Tweet

Rob: Outside of the data assets, governments and everything that surrounds the concept of data, what should banks that have not explored these technologies look at for?

Peter: The data one is the big one. There’s other things like integrations with core systems and this is one that non-bank industries have to deal with. Everyone has core technologies and often times legacy systems. Banks need to have two things. The first is banks tend to have more of them and there often necessary to any bank initiative. Number two they have regulatory pressures that make integrations with backend systems more complicated and with higher risks. The phrase “start small” is perhaps overused, but it’s not untrue. That’s why a lot of the leading banks that have bots out there started small, where they could control a little more of the regulatory systems and were plugging into two legacy systems and four hundred. One more thing I’ll say that’s not surprising, but maybe they won’t think about much is internal politics. Banks tend to still be organized, but in lines of businesses there will be clashes. What does this bot do is also a question of who owns it. That’s already a question we’re hearing bank executives start to ask and sometimes answer, while other times not be able to answer.

The phrase 'start small' is perhaps overused, but it’s not untrue. @P_Wannemacher #banking #fintech Click To Tweet

Rob: Here at Abe AI, we think of the way the consumers look at the bank and the way that the bank looks at the consumer. Trying to own one initiative across a couple silos is an amazing initiative, but to think that you can create one generalist touch point from consumer to bank across all eighty-eight silos is going to have to be a serious initiative for the institutions. I’ve acquainted to the way that they huddled together and created apps. It’s such an interesting parallel between the way that apps developed and the way that the data layers around apps developed. Many people say that there’s not too much difference in the way that apps consume data to the way bots should consume data. I don’t necessarily disagree with that concept.

Peter: In broad strokes that is absolutely true. Just like with apps, it will not be a sufficient answer to your customer to say that it’s not your line of business. I mean that in a literal sense, but I also mean in terms of designing the experiences, sort of saying that’s not your problem won’t work. You see with the leading companies like USAA, banks in Poland, all the Australian banks, and CIBC in Canada that the answer of it not being your problem is increasingly unacceptable and that’s a good thing.

Rob: We’ve talked about some of the big banks, but how do you think the small and medium banks should be differentiating their AI strategies from these larger institutions. Is it open-source that they should be taking advantage and trying to create some of things in their own internal innovation teams, or should they be trying to think about it the same way the larger institutions are? Going back to the parallel with apps and see how larger institutions lead the wave of apps and were highly successful with drawing customers in and then you have the other institutions that caught up to that movement. Can you parallel to that for me?

#AI is going to change the way bank customers do things and the way banks do things. @P_Wannemacher Click To Tweet

Peter: There are going to be differences in strategy, execution, investments, and a whole number of things. I do not believe there will be major differences in how AI affects their business, which isn’t always true. In this case, AI is going to change the way bank customers do things, change the way banks do things, and that’s will be as true for the small banks as it will for the big banks. The threat of not acting quickly or appropriately is just as serious for small banks, to medium sized banks. It’ll differ in how quickly it might happen, but it won’t change the degree of impact in terms of how they should. I do believe it is very unlikely that a bank, of any size, will be the organization that develops the breakthrough AI system that changes everything. The partnership component of this is important. A friend of mine has a really great story on the evolution of innovation. There’s the old school secretive labs that no one knows about where they are doing R&D. Then they moved into shops that everyone knew about, but no one was really exposed to and then it became solely about partnership with fintechs. It may be too often said at conferences, but there’s a reason for it. Fintechs belong in the banking ecosystems.

#Fintechs belong in the banking ecosystems. @P_Wannemacher #banking Click To Tweet

Peter: I do think that open source and partnerships are the way to go for small and medium sized banks. I actually think big banks need to be careful. I think big ones would actually be wise to partner. We see this with the Google’s and IBM’s most obviously. There are sometimes where you need to pick. Macquarie Bank in Australia, this is public so I don’t mind revealing this, in their retail efforts decided that they need to carefully choose three to four areas where it made the most sense for them to build their own IP and technology, rather than go out and do what someone else is doing well. Most of the time they were openly happy to use what other people were doing well. Other times, banks want to build it themselves because of this old problem of internal political reasons. I think Macquarie Bank is a little smarter with identifying places where there’s enough value of having their own IP. For them it’s interesting because some are unsurprising like marketing and search that they decided to build on their own, but machine learning is interesting. Machine learning is, in Forrester’s research, is a subcategory of AI, or a subset of AI enabled technologies. Macquarie Bank wanted to build their own machine learning algorithms that was essential, in their view, to their business. It will be years before we know if that was the right move, but I think that banks of all sizes need to be careful about what they’re choosing. The smaller banks will have fewer areas that they build themselves and the bigger banks will have more, but I think for all of them there’s a pretty similar approach. It’s just a matter of how much money.

Open source and partnerships are the way to go for small and medium sized banks. @P_Wannemacher Click To Tweet

Rob: I was on a panel in New York at a Lending Conference and one of the questions that came from the audience was, “As a bank how do you even begin to acquire the talent that it takes to create these not so rudimentary machine learning algorithms?” As a statistician, I’m sure you can identify very well what the difference between the rudimentary algorithms that you get off the shelf or GitHub, and where an artist can come in and start painting with math and apply mathematics. Can you speak on what banks could be doing different from a culture and recruitment perspective to get this?

Peter: I’ll say it in two ways. One gets back to the answer I gave last time, which is partnerships. There are brilliant fintech that are good at hovering up brilliant talent working with them is a great idea. I was talking to a banking executive and he gave some great advice. He was saying that for the past five years we’ve been trying to get the best development talent in engineering, developers, and so forth. We get the best talent out of the best universities, we’re willing to pay top dollar, we’re willing to throw lots of bean bag chairs, bowling alleys, and free lunches at the problem. While we will not stop doing some of that, we recognize something in the past year and that’s that we won’t win that war. Instead what we need to be willing to is hire good and train to great and I think it’s the only option for some of those banks. You can hire a lot of great talent, I don’t mean to give up on going out to Silicon Valley and talking to those folks, but there needs to be a recognition that the top five draft picks are not always the best players down the road.

You need to be willing to hire good and train to great. @P_Wannemacher #fintech #banking Click To Tweet

I’m not a corporate culture expert, but I will say this, the banks that I see that have the best talent and the best training of talent tend to have a culture where there’s an acknowledgement that what they are doing isn’t especially “sexy,” but what they are trying to do is really bold. I know a bank that has a goal of being the number one mobile bank in the world. That may not excite every 17 year old or 22 year old coming out of college, but that’s a fun project to be on and a fun goal to have. You can’t just say that and not be willing to throw money behind it, but I think that’s what we are seeing as being successful. The last thing I’ll say is that organization matters. Again, I’m not the expert on organization structures, but firms like Vanguard, USAA, companies that are not brand new, and those that aren’t new banking or financial services companies, have an organization built around customer journeys.  Forrester wrote a report years ago called, “One Customer, One Organization, One P&L” and the idea was not that any company would actually get there, but that should be the goal. You should be trying to get as close to that as possible. I think that doesn’t immediately change the culture, but it starts to juice those processes and people to start and get some where.

Rob: I think some of the banks really underestimate the assets they really have in this scenario. From my experience with these individuals that are highly talented in advanced machine learning, particularly deep learning, the technology that they create requires enormous amounts of data. Most startups don’t start with a load of data. For a data scientist are you going to want to go optimize how quickly your Uber gets to your front door or would you rather help a person understand how to manage their debt better in relation to the rest of the world.

Peter: You’re absolutely right. Banks have always been really good at collecting data.  They are also pretty good at storing and managing data. One piece of evidence is that trust in banks remains pretty high. People don’t love banks. It’s a little bit like congress. They tend to dislike banking, but they kind of like their bank.  As a data point of this is, in the U.S. bank customers trust their bank more today than they did before the great recession. That’s not all about data, but it’s tied up in that relationship. I think you’re absolutely right and those data assets are enormous. We saw a big bank that spent about four and half years completely rebuilding their data infrastructure so that they could enable things that they could never think of doing today. Things from real time predictive analytics and real time transactions for customers. If that is the base layer of great AI with a potentially great bot experience on top, that’s awesome and is why I’m bullish about this.

People tend to dislike #banking, but they kind of like their bank. @P_Wannemacher #fintech Click To Tweet

Rob: We’re halfway through 2017, what do you think are going to be the big emerging fintech trends for 2018?

Peter:  Fintech has done a good job in making things fun and covering a fair amount of ground. I think that broadly speaking, fintech has done a pretty good job of covering a wide range of customer journeys and people’s financial needs. I think that fintech will continue to be broad in its scope of what the trends are.

#Fintech has done a good job in making things fun and covering a fair amount of ground. @P_Wannemacher Click To Tweet

There are two areas of financial services that I think are undervalued at this time by both big providers and fintech, but they overlap a little. The first is life events. From our research we know that for more than twenty-years the number one driver of financial product purchasing, of any kind and from any kind of company,  are life events and those tends to be the big demographic shifts, marital shifts, career shifts. Examples would getting married, getting engaged, getting divorced, getting separated, having a child, adopting and so on. There is roughly a dozen or so of them that are the main catalyst for people going out and looking for financial products and accounts. Almost all banks have life events buried somewhere on their marketing website.

#Fintech’s done a good job of covering customer journeys and people's financial needs. @P_Wannemacher Click To Tweet

Let’s talk about realtors as an example. If you’ve ever bought a house then you know that you develop a really close relationship with that realtor, even though you only spend six months with them. After that, that’s the end of the relationship with you. I wonder if there’s some sort of fintech that will start to play the role of realtor for financial services. All financial services are built on this idea of being your go-to support for twenty plus years, but I wonder if there will be any smart fintechs that will not do that and be there for you during the pregnancy finances, or during a move. I am not smart enough to come up with that company, but I think that others will start to do that.

The second one is a concept Forrester calls “shared finances.” We didn’t make up the term, but I will quickly give you our definition, which we did create. Shared finances means any situation in which one person acts as an observer of, a partner in, or a proxy for another person’s finances. Each time I say this somewhere with bank audiences, there’s always a banker who says, “Oh, you mean a joint checking account, or credit cards with multiple card holders.” And I say, “Yes, those fall into the broader umbrella of shared finances.” Those aren’t doing anything close to covering the range, the value, or the complexity of people’s major personal financial lives. Some examples include divorced couples who are co-funding a child’s education, or business owners and partnerships that are relatively new.

It’s easy to manage finances when it’s a father and son running a business, but even that can get complicated. It is much harder when it’s a group of friends, even with small things like friends going on vacation.  I wrote a blog post about called, “Splitwise is a Fintech Disruptor That Shows The Potential of Shared Finances.” The blog is about a small startup, Splitwise, that has an app that makes it easier to manage what everyone is spending on vacation, without having to write it all down. As parents get older, the adult children at times manage their parents finances. There’s a fintech company called “Tomorrow” that deals with the financial implications of death and that’s a dark end on this questions, but there’s a lot of need there. Banks have left that area pretty open and as a result I think it will be open to disruption. I think fintech will keep doing what it’s doing, which is really good, like in Jurassic Park where the Tyrannosaurus Rex is really good at testing wires, that’s what fintech is doing all the time. Fintechs will try something out and if that doesn’t work they’ll try something else out. This is something the banks have never been good at and fintechs are really good at. I have no doubt that bots will be a part of that.

Fintechs will try something out and if that doesn’t work they’ll try something else out. Click To Tweet

Rob: For all you data scientists out there, tap into the transactional data and start predicting life events. That sounds like an excellent use of AI.

Peter: Absolutely!

Rob: This has been an excellent session with Peter Wannemacher from Forrester Research on, “Are Bots Ready to be Bankers?” I hope you all enjoyed the content. Thank you so much for joining us Peter!

Peter: You’re welcome! Have a great day!