Autonomous finance uses AI to make financial decisions on behalf of consumers without the need for direct human input. The service has become especially relevant over the last year as consumers have struggled to maintain financial health during the COVID-19 pandemic. In this episode, Paul Condra, head of emerging technology research, and Robert Le, senior emerging tech analyst, discuss how autonomous finance helps consumers better manage their financial health and performance, as well as the challenges for the technology—including computing costs, consumer trust, regulations and transaction categorization.
Listen to all of Season 3 and subscribe to get future episodes of "In Visible Capital" on Apple Podcasts, Spotify, Google Podcasts or wherever you listen. For inquiries, please contact us at email@example.com.
Watch the webinar and download the presentation here. Transcript Adam Lewis: Welcome back to "In Visible Capital," a show that discusses the inner workings of the private markets. I'm Adam Lewis, a private equity reporter for PitchBook News.
Alexander Davis: I'm Alexander Davis, editor-in-chief of PitchBook News. Today, we'll be sharing a fascinating conversation on autonomous finance from a recent webinar with Paul Condra, our head of emerging tech research and Robert Le, a senior emerging tech analyst who focuses on fintech and insurtech.
Adam: Alec, would you believe it if I told you that you could purchase a robot to run your personal finances and wealth management?
Alexander: Well, normally, Adam, the skeptic in me would say that that's probably just a little impossible-sounding. However, it's not about me. The Silicon Valley fintech mavens, you never know what they're going to come up with. The fact is that millions of dollars of venture capital are being bet on apps that can do all of those things and more. I think that the conversation between Robert and Paul really illuminated what's at stake here and what the issues are that people need to know about.
Adam: They specifically cover how autonomous finance can help consumers better manage their financial health and performance and keep people like me from converting their 401(k) to, say, Dogecoin or another fringe cryptocurrency. They also discuss the pathways to achieving full autonomous finance and challenges for the technology, including computing costs, consumer trust, and regulations. Let's jump right into the conversation with Robert.
Robert Le: We've recently done research in autonomous finance, and I'll just explain what it is and continue in deeper into some other different areas of it. We're currently focused on the retail side. There's also an enterprise side of the application and we're looking to do research in that area in the future, but we define autonomous finance as applications that use AI, machine learning and other automation algorithms to make financial decisions for customers without needing the direct input.
You can think of how autonomous driving can speed up a vehicle, brake, change lanes. Autonomous finance can automatically pay bills, move money into higher-yielding assets, pay down debts based on interest rates, or find better insurance policies. Really, there are numerous use cases out there for the technology. Consumers are struggling with personal finances. There's a lot of research on the psychology side that shows that humans innately are not developed to manage finances well or to save for the future. A recent survey conducted last fall saw that 64% of millennials were living paycheck-to-paycheck and 82% could not afford a $500 emergency expense.
Consumers generally make a lot of poor financial decisions, whether it's getting into too much debt, overspending or making bad investments. Some of this can point to the lack of financial literacy, but also back to the psychology aspect of how we view money as well. For instance, there's a suboptimal financial behavior called the credit card debt puzzle, which we talked about in our autonomous finance report. It's the tendency to concurrently keep cash, which is like low-yielding in a low-yielding asset like a savings account and also, at the same time, holding high-interest debt like credit cards.
It'll make more sense for you, for a customer to use that cash, to reduce that, but really, 30% of US households do not do that. You look at just banking fees, consumers pay more than $10 billion a year in overdraft and non-sufficient fund fees, ATM fees and account maintenance fees; [it] is a huge source of revenue for banks. The other challenge in consumer finance is just budgeting in general is very onerous. It's highly manual. The tools available on the market range from spreadsheets like Excel to applications like Mint or You Need a Budget, which are products that have been on the market for almost a couple of decades and have seen very little innovation since inception.
You Need a Budget and Mint were founded in 2004 and 2006, respectively. Quicken was really the OG personal financial management software developed in the '80s. Quicken is pretty clunky. It's feature-rich, but many consumers today, especially younger generations will likely not understand and use most of those features. Credit Karma, which came out in 2007, it's a decent PFM software, but it only focuses on savings, credit and loans. Nothing on the investments or other financial products.
Then we've seen a lot of consolidation in the space as well. In the past few years, you see Intuit has acquired two of the prominent PFM players who I just mentioned, Mint and Credit Karma. What we think is important for a PFM software is that it has to take a holistic approach to deliver the most value to the customer. We think autonomous finance will take that holistic approach.
Paul: Robert, I have one question for you on this evolution. When I think of automated finance, I think of like robo-advisers and that whole kind of automated wealth management part of the market, but you haven't really included that here. Are you focused more on a different aspect of finance automation? Is it just more on the budgeting personal management, not so much on the financial investing side?
Robert: Yes. The way we see it is PFM is more holistic, so it includes robo. Robo is just one use case, one aspect of personal financial management. What we're just highlighting here is what are the more broader applications? Robos could plug in anywhere in this space.
Paul: Got it. Also one other question, the stats you mentioned that we're focused on the millennial demographic. Who's the target demographic here that you see for these kinds of solutions that are coming to market?
Robert: Definitely, the millennial and Gen Z. I think the younger generations, as they start using the natural products, they're more likely to use a mobile application, go online and use digital forms of financial services rather than the older generations, which are more in-person and branch. Autonomous finance will obviously be in a digital form. That's going to be the generation that will most likely use it. Also, I think just financial literacy, it's a lot lower in the younger generation. It just makes sense for them to use autonomous finance where there's a lot more hand-holding along of "how do you manage your finances?" than the other generations.
Once you highlight some of the companies that are doing some interesting work within autonomous finance, Wealthfront is the first one that ... They first entered the market as a robo-adviser. They recently came out with a product called Autopilot, which is their savings automation product that constantly monitors an external checking account or the Wealthfront cash account. It automatically moves excess cash into various designated investment accounts at Wealthfront.
Right now, that excess cash is based on a threshold level designated by the customer. We think that in the future, Wealthfront can automatically determine that threshold based on a customer's spending behavior.
Douugh is an Australian neobank. They also have an automated money management feature that is integrated directly into the banking app. It uses AI models to analyze transactions, to evaluate patterns, and they're able to detect anomalies and forecast future account balances for customers. It also can detect income and bills and automatically provision money for bills by sweeping the extra cash into savings, and then later come back and pay for those bills.
Personetics is an interesting company. They provide more of the enablement technology for autonomous finance. That technology is able to analyze financial data in real time to understand the customer's financial behaviors and automatically move money into savings accounts, investments or pay off bills.
Their technology right now is being deployed live with a couple of retail banks and directly into their banking apps. They're helping these banks build financial wellness programs, allowing banks to offer financial insight and advice to the customer, but also cross-sell relevant personal lives, financial products as well.
Last one is Curve. It's an application that aggregates all of the consumer's bank accounts and credit cards into a single wallet.
Just this week, they introduced a suite of autopilot products that provide personal lives and automated money management, includes what they called an anti-embarrassment mode in which if you have a default card and it's declined at the point of sale, it automatically routes that payment to a backup card without the customer having to pull out a card or change the card seat to run the transaction again. It automatically just switches that.
They also have a rewards auto-spend feature, which typically, you look at rewards for credit cards, you have to manually request the cashback or the reward amount. Well, Curve will automatically withdraw from your rewards balance for purchases if that balance is sufficient to cover a purchase.
Paul: It sounds like for a lot of these, it requires the user to connect a lot of different—or share a lot of different—data sources with the app or with the startup. Is that kind of an obstacle? How are these startups overcoming that in terms of getting out this information that they need from the users?
Robert: For like Wealthfront, if you connect an external account, they disconnect through Plaid. Plaid is a pretty common one. Curve, I'm not sure which aggregator they use. They're based in Europe. It could be Plaid, could be Tink, but the same thing. You just sign on to your account and then it'll automatically make that connection through an API.
Paul: Do you have any sense of ... We have digital private companies, but I'm wondering if you have any sense of adoption uptake or growth. Are these companies raising a lot of money?
Robert: Yes, they definitely all raised a significant amount of money in the millions, tens of millions. Wealthfront have been one of the larger robo-advisers in the US. They have over $15 billion of assets under management. That's really still their core business, but what they're saying is that they can expand their core business and go into other areas of consumer financial services, but these are pretty large. They've all raised tens of millions of dollars.
The way we look at it is autonomous finance will likely exist under two models. It can exist under the centralized model in which an established fintech company or traditional retail bank will develop autonomous finance that they will lay ... on top of the existing products and embed within their financial applications. We believe companies most likely currently developing autonomous finance solutions right now are the large neobanks like Chime and Revolut, as well as the non-bank players like PayPal and Square's Cash App. In a decentralized model, autonomous finance will exist as features of a separate standalone wallet application.
Curve, as I just mentioned earlier, they run on a decentralized model. They're now developing autonomous finance. Curve customers, they do not open any financial accounts directly with Curve, but they connect it all into a single wallet application.
Paul: It sounds like just on the surface, if you're this kind of decentralized app that's connecting financial services on the back end, you're going to run into certain functionality walls because you're reliant on whatever APIs are available to you relative to the centralized model where you can provide whatever service you want because it's within the proprietary suite of financial services you're already using. That stand out to you as kind of a barrier. Does that make it seem more likely that one model's going to be more successful than the other?
Robert: Yes, definitely. I think what you mentioned, the centralized model is just more attractive. In our recent research, we even talk about what we think the market value could be for that. Definitely, it just makes more sense for everything to be centralized, but I think the decentralized model, it's lighter. The barriers to entry is a lot lower. If you just look at Curve's model, they don't really own or have any of the assets like deposits or anything on the balance sheet. It's just a lighter model and it's easier to develop. I agree. I think the centralized model will deliver the most value to the customer.
Paul: I'm also curious how regulation works in here where you would have a situation where an algorithm is making financial decisions on behalf of the customer. At what point does that kind of cross over into a regulated space where you have like an automated-financial advisor relationship? Is this a focused area of regulators or it is still sort of a gray area?
Robert: It's definitely still a gray area. We'll talk about regulations a little bit later on the presentation as well, but right now, there really isn't any regulatory structure in place for even a lot of fintechs in general. You look here in the US, even a neobank like Chime or SoFi—SoFi, now that they have a bank license, [is] different—but like a Chime, they're not really directly regulated by a financial regulator because, well they're not really considered a real bank. A lot of these fintechs do operate in this gray area right now. There are a few key adoption hurdles. The first one is the AI computing cost. You can think of AI development, computing power and data storage, all that is really expensive to deploy.
Also, just related, underlying financial infrastructure and core systems will likely needed to be updated as well. Autonomous finance requires real-time access and movement of money. Outdated core technology and banking ledgers, those will likely need to be updated too. It's not sufficient to scale this technology.
We think that gaining consumer trust, which is imperative with any financial service, especially under autonomous finance where a customer is asked to pass the reins of financial management to basically a computer program. There would need to be a certain level of trust established before that can happen. Positive viewpoint is that there is a recent survey that was conducted and it showed that 67% of consumers trust a robot more than a human to manage their finances. We think that a lot of the ship has been driven by the success of robo-advisers. Is this a positive development for autonomous finance? For sure.
Paul: It's interesting to think about how the consumer trust just goes farther and farther over time. I remember when I think maybe it was Coinbase, when I first opened the Coinbase app and I had to take a picture of my ID in order to get verified or approved, and the thought of taking a photo of your driver's license and sending it to an app was unheard of, and it's viewed as this huge, tremendous risk, I think from a consumer trust perspective. I'm not going to send a picture of my ID to some app that I don't know much about. Now, it just seems like, "All right, here you go. Snap, snap." Do you know what I mean? I'm sending it off to any app; I don't even think about it.
Robert: Yes, it keeps shifting away. You think even when Amazon ... Amazon was the first retailer that allowed you to save credit cards. When you go on and make a purchase again, you don't have to enter your credit card details. Back then, before Amazon, no one did it. You have to enter it in each time. Everyone's like, "Why would you save your credit card online? That's so risky." Now, it's like you don't even think about it when you're entering a credit card details. It's definitely shifting.
Paul: Yes. It's just like a further abstraction away from those core financial consumer practices that we use to follow from a risk-management perspective. Now, we're just much farther away from that. I think that makes it easier potentially to sell in new products and services, like a much more abstracted, digital financial layer on top of what was normal 10 years ago.
Robert: I think this is a favorable for our time as finance. There are some risks, especially if you're automating into investments. I think at the end of the day, consumers really value customer experience and efficiency over some of these risky or seems like risky behaviors in the past, but they're not.
Paul: Part of it is you have to do it. If you want to participate in it, you have to do it. I don't think there a crypto app you can download and use without sending a picture of an ID. I don't think there's really any other way. There's exceptions, but there's these corridors have been made and you have to pass through them if you want to access these new services.
Robert: Yes, absolutely. Regulation is definitely a key adoption hurdle. There's numerous financial and data privacy regulations that you have to comply with. Those regulations are constantly changing as well. There isn't really any financial regulatory structure in place for autonomous finance right now because these algorithms and AI at times operate in a black box. We think that explainable AI solutions will be needed to show regulators that these technologies are in compliance with the various laws. At times, finance can also potentially introduce new systemic risks into the financial system.
Similar to how high-frequency trading can lead to [...] Or accelerate equity selloffs or withdrawal of deposits can be accelerated under certain scenarios in autonomous finance. It just increase liquidity risks to some of the banks. I think that's something that regulators will definitely have to take a look at as well.
Paul: It definitely seems like very early ground for putting some regulatory compliance framework over AI. I think I read that they've been making some progress on that in Europe with some new AI framework regulations, but it certainly seems like the US is not close to establishing something to govern how these algorithms work.
Robert: Yes, absolutely. The regulators in Europe and in the UK are definitely more progressive than here in the US. The last hurdle is data integrity. If you look at just financial data categorization today, like through some of the major financial data, aggregators still is not 100% accurate. For instance, if you and your friends split a meal at a restaurant and your friend pays you through Venmo, a lot of these aggregators will count that as income. That's just not true. A lot of this financial data is not granular as well.
Let's say you go to a gas station and you purchase, I don't know, a food item there, it will categorize that as a gas purchase and not a food purchase. We need more granular data as well for autonomous finance to really work well. Now, still a huge challenge for some of the financial aggregators today like Plaid, Tink and MX, so we think there needs to be significant improvement in this area before autonomous finance can truly develop.
Just looking forward, what we think is going to be next for autonomous finance, the next logical step is to develop and integrate bank accounts with payroll providers. There are a lot of startups like Pinwheel, Atomic and Finch who are developing payroll APIs. Plaid, I believe, is developing an API for payroll access as well, and this allows for integrations with payroll providers like ADP or paychecks. We've already seen a lot of the early wage access programs from neobanks and even incumbent like Capital One which allows you to access your paycheck two days early, but we think that by directly integrating with payroll providers, people could access their paycheck in near real-time.
At the end of each working day, autonomous finance works really well with real-time data. It can move some of the daily earnings into a spending account and the rest into a savings account or investments. Just imagine like at the end of each working day, some of that capital automatically gets invested instead of having to wait two weeks later for you to get some of that cash to make investments.
It can also be levers, some kind of income smoothing for non-salaried workers. Eventually, we think there'll be automation across all aspects of the consumers' finances. Bills and expenses will automatically be paid using the best type of payment account. For instance, I have an Amex card that reimburses me $15 a month for Uber rides, but my Chase card gives me the highest rewards for Uber. I kind of have to remember to toggle between the two cards when I use Uber each month. I just think that should be automated.
I think credit cards will likely change or be delivered as a cash flow as a service based on real-time financial factors such as up-to-date payroll and daily spending, and they'll be underwritten at the transaction level. Let's say I make a purchase for a cell phone, autonomous finance can determine if I get any rewards for it. If so, the funds will come out of that account. In another scenario, let's say I don't have enough funds in my checking account to fully cover for that cell phone, it can cover up— That purchase can be partially funded by a credit line and then the application just notified me that it'll cost me X amount and interest in the next three months to make this purchase.
Paul: It's interesting because when you checkout online, you have a lot of options. You have like buy now, pay later, you have your credit card options, maybe you can pay with your bank account, maybe you can do COD. Now, you can't do that anymore. There's all these options that are available to you and as the consumer, you optimize that at checkout. You optimize your purchasing behavior based on what looks the most attractive what you think is the best way to buy something.
You're almost suggesting that that process, to the extent that the platform you're using knows you almost becomes like a little bit of a marketplace where those cash or liquidity providers can compete for your attention and serve up what they think the best solution is. It could be a little bit of lending, it could have been a little cash, it could be this rewards card, maybe you have these kinds of blended transactions that are just fully automated. Is that what you're talking about?
Robert: Yes, that's exactly what I'm talking about because right now, when you make a purchase, it's typically through only one account, but you can just split it up into multiple accounts based on interest rates, based on rewards, based on how much cash you have as well. I mean, all that can easily be automated and isn't really an application. Now, that does and we think autonomous finance can.
Paul: When you think about— Talking about the outlook, is there space here for a startup to really bring it a revolutionary type of model and scale, or is this more about incremental improvements and we're going to see the large incumbents? You mentioned PayPal, Square, even maybe like a JP Morgan or Bank of America, they're going to be the ones that are either maybe buying these vendors or bringing this technology to market and scaling it through those existing platforms. Do you think there's room for the new disruptors to grow?
Robert: There's multiple angles where this is going to be developed. One way that we're seeing it is within the larger neobanks and with the fintechs like PayPal and Cash App. They'll develop and they'll develop for their own ecosystem. The retail banks, the traditional banks are also developing on their own. There's also companies like Personetics who is developing the technology to sell it. They're not going to be using it for their own customers. They're selling it to banks and other financial institutions.
Then there's the aggregators like Curve that I mentioned that is really asset light model for them to just develop it on their own. They just connect all the accounts into one. Overall, it's still a really technically challenging technology to develop and it requires a lot of data. I don't think there's going to be five or 10 awesome autonomous finance solutions out there in the market. I really think there's only going to be a feeling, eventually it's going to consolidate around the few providers of that.
Paul: I think at this point, currently, a lot of these apps ask consumer in terms of connecting accounts, connecting cards, whatever it may be. There's almost an activity of always ensuring that when you're using an app aggregator, that your connections are up-to-date, your passwords are working so that you're actually pulling the real-time data, it seems like there's still just a lot of— Just in using like a financial management tool, there's still a lot of friction. It would seem like that stuff would also have to get ironed out a little better than it is currently.
Robert: No, absolutely. That's what I mentioned earlier. That's one of the challenges, is the financial aggregators like Plan, they have developed a product that has really decreased the frictions in connecting accounts, but there's still a lot of challenges. Connections get dropped a lot, categorizing transactions aren't accurate. There's still room for improvement there and it's definitely going to require for autonomous finance to really develop.
Paul: Another question I had was different direction because we're talking a lot about DeFi these days and there's a lot of activity there. When I think about things running autonomously, automated services, it really seems like those kinds of networks are potentially a really good place for this product to exist because it's just open and you could develop products that can work across these networks and do things much more easily than having to connect to a bank's internal platform. How does that fit in here?
I know this isn't really the focus of what you're talking about, but do you think about that being integral to the growth of autonomous finance or does that feel like it's a little bit separate at this point?
Robert: When we talk about the decentralized models earlier, it makes sense for us to think about DeFi. For those of you in the audience that don't know, DeFi is decentralized finance using blockchain. I think autonomous finance and DeFi can and will likely coexist. DeFi is essentially taking out the middleman in the financial system. Whether it's the credit card networks and payments or banks, savings and loans, or the brokerages, trading and investments, in theory, if a consumer was to conduct all their financial services through DeFi applications and protocols, there's still an opportunity to personalize that service to a specific customer.
You think about right now, the automation and DeFi is through smart contracts, but that automation is focused on facilitation and processes of moving capital, not on individual needs and unique circumstances. I think that's where autonomous finance can be layered in enabling personalization really at the individual level.
Paul: Interesting. I want to thank Robert for joining us today to talk about autonomous finance.
Alexander: Well, that does it for this week's episode of PitchBook's "In Visible Capital" podcast. Thanks for listening. For more information, including links to the report and webinar covered in today's episode, you can visit pitchbook.com/podcast. Please rate and review the show to help others discover and join us next week for a great discussion about more emerging tech issues, specifically, we'll be focusing on property tech as well as mobility and a really fascinating conversation about air taxis. Until next time, I'm Alexander Davis.
Adam: I'm Adam Lewis. Thanks for listening.