Financial dashboards like Mint [acq. by Intuit 2009 / $170M], Clarity Money [acq. by Goldman Sachs 2018 / $100M], and our own Guided Portfolio Summary allow users to get a single view of their financial data, do some basic calculations (e.g. determine net worth), and perhaps consolidate that data in a single score to measure their progress towards goals like a secure retirement. However, the cognitive load to process the information and act has remained entirely with the user: Once you know your total debt or your retirement readiness score, it’s up to you to determine and take the next step.
But that won’t be the case for long. Early stage investors, and some innovative incumbents, are betting on autonomous finance—AI-driven financial services that independently make decisions or take actions on a customer’s behalf.
Credit Karma has been one of the early leaders in autonomous finance. They don’t stop at just calculating and displaying your FICO score and offering suggestions for how to improve it. Instead, Credit Karma also recommends credit cards or loans with the highest probability of being approved and enables a user to run a pre-approval process against them all with a single click. While not fully autonomous, Credit Karma has advanced well beyond offering one-to-many recommendations that users have to then act on—which probably explains much of why Intuit paid $7.1 billion to buy them in February.
The Three Characteristics of Autonomous Finance
What distinguishes an autonomous finance service from other digital offerings?
- The ability to source external user data and combine it with internal data, in real-time and at massive scale. In autonomous finance, the goal is not to deliver a generic or overall best-rated product, but instead one that is built around the full needs of a specific customer in a specific moment. Take Digit: it connects to the user’s checking, savings, and credit card accounts, builds a predictive model of earnings and expenses, and calculates the amount that is “safe to save” every single day. It goes beyond the “round-up” (i.e. save the change) or even “found money” (fixed percentage of income or expenses) offerings that have become commonplace. Digit offers user-specific savings maximization in real-time. To get a sense of how much data it takes to deliver autonomous finance: Every day Credit Karma collects 2,500 datapoints for each user.
- No human intervention is required. By default, a service isn’t autonomous if it requires the user or a third party to take action. Northstar Money, for example, doesn’t just build income and spending models for a user. Instead, with each paycheck Northstar directs the allocation of those funds, putting some towards paying recurring bills, some towards a short-term goal like a vacation, some towards retirement, and the remainder for discretionary spending. However, just because human intervention is not required, doesn’t mean it couldn’t be offered, perhaps as an optional value-add service. Northstar offers a premium option where users can speak with an advisor, who can help them adjust goals and automation criteria.
- An action orientation. Autonomous finance services don’t just alert or inform the user of an upcoming bill or missed goal. Instead, they suggest specific actions and, in most cases, empower the user to automatically execute them. Albert’s Genius functionality, for instance, monitors your checking account and, if it notices a shortfall, automatically issues a 0% interest $100 pay-day loan. Behaviorally, autonomous finance generally makes users feel good about reaching goals (“Congrats, you have saved $257.34 this week! Way to go!”), as opposed to feeling guilty because they can’t stay within a budget.
Technology Advances Have Set the Stage for Autonomous Finance
Three main developments have converged and created the right environment to advance autonomous finance:
- Systems of record and engagement. Over the past twenty years, we’ve seen the widespread sourcing and collection of user data—everything from credit card spending to health records, emails, and online behavior. In the financial services space, these data have spurred the creation of dashboards, calculators, and other planning tools designed to engage the customer by helping them better understand their financial picture, determine their progress towards goals, and choose products.
- Artificial intelligence. Using artificial intelligence, routine cognitive functions (i.e. situations that "I’ve seen before") are relatively low on computational load. Credit Karma’s 2,500 datapoints per individual per day is a dataset that a computer can process better than a human brain, which instead excels at non-routine cognitive functions ("I’ve never seen this before"). Many financial decisions, even some very complex ones, are made based on routine cognitive (mathematical or statistical) functions, and so are the easiest to tackle using AI: think of rebalancing or tax-loss harvesting an investment portfolio or calculating which trades from across all your household accounts will minimize the tax impact of a large cash withdrawal. And with AI advancing in non-routine cognitive functions (reinforcement learning and continual learning), autonomous finance services are set to expand even further.
- Commoditization of core financial services. In the past few years, financial capabilities like payments, KYC, fraud prevention / AML, custody, clearing, and financial data aggregation have become available off-the-shelf from companies like Plaid, Onfido, Galileo, and Apex. Embedding these functions in your services requires little more than a pay-per-use API.
With core financial services capabilities commoditized, fintechs have focused on higher-value offerings, like Scalable Capital [founded 2014 / $81M total funding / $2.3B AUM] with volatility prediction of individual assets in real-time, or Just Invest [founded 2018 / $2.1M total funding] with daily rebalancing and tax-loss harvesting via direct indexing. These offerings are the realm of autonomous finance. For new fintech founders, autonomous finance offers the opportunity to marry the zero-marginal cost of software with high value services, the same services that high net worth individuals have historically outsourced to expensive accountants, financial advisors, investment advisors, and tax attorneys. As such, the potential unit economics of these services make an appealing investment thesis for VCs.
Mapping Autonomous Finance
How advanced is the market for autonomous finance in 2020? Where are fintechs and incumbents making their early bets? A good way to map autonomous finance services is along two axes—the Level of Autonomy inherent to that service and the Breadth of Advice it offers.
- Level of Autonomy. On one extreme of this axis is the realm of the dashboards (systems of engagement), while on the other extreme is full autonomous execution. Clarity Money is an example of a dashboard that highlights the customer’s recurring subscriptions; however, it’s up to the customer to decide which ones are not needed and manually cancel them. Tally, on the other end of the axis, aggregates a customer’s credit card data and analyzes interest rates, minimum payments, and due dates to build a customer-specific debt repayment plan. Once the customer connects all their credit cards and their checking account, Tally regularly withdraws an optimized amount of money (buffered by a Tally low-interest loan) and distributes payments at the right time to the different credit card providers.
- Breadth of Advice. On one side of this axis we have autonomous finance services that target a single, typically narrow, financial need. At the other end, we have services that serve a wide range of financial needs. Tally and Digit are examples of autonomous finance apps that focus on a single financial need (debt and savings respectively). Acorns, on the other hand, covers a broad range of financial jobs to be done. Daily, Acorns can move money from the user’s checking account to their savings account, automatically invest long-term savings in customizable autonomous portfolios, and direct retirement savings into retirement-specific accounts and autonomous portfolios.
Offerings in the top right quadrant of the map fall under what Ernst & Young calls Personal Finance Operating Systems (PFOS). Due to the advisory nature of some of their services, PFOS providers like Acorns and Stash will face higher regulatory burdens. They’ll also need to engender deeper levels of customer trust to pull off their "all eggs in one basket" approach. An alternative PFOS approach would be to have these services delivered by a third party on top of regulated accounts held at other financial institutions. Astra Routines, for example, allows a user to define "If This, Then That" rules across financial services providers—something like, "If the balance on my Chase checking account goes over $2,000, sweep the difference into my Fidelity brokerage account." The company is very young and the routines are still limited, but the strategy is clear.