Non intuitive use case for Banking as a Service: Autonomous Finance
- After the advent of digital, banking is still, at its core, a ledger.
- Autonomous finance offers a reimagined idea of banking.

The following is an excerpt from our recently published report on non intuitive and interesting use cases for BaaS. Our Outlier members have access to the full report.
Other than making the customer experience better, easier and more enjoyable, banking has not significantly progressed in the digital age.
The same products and services still exist. Banks still offer similar checking, savings, multi-currency accounts, and money transfers -- now, it’s just with a nicer frontend. At its core, banking is and has always been simply a ledger.
What is autonomous finance
- Autonomous finance is a new way of thinking about what can be done with banking.
- Using AI and Machine Learning, autonomous finance can remove human input from decision making, enabling every customer to have a virtual private banker making all their financial decisions for them.
- Although typically associated with investing companies like Betterment and Wealthfront, autonomous finance has tremendous potential in many other areas of finance, like banking and lending.
While companies like Qapital, Chime, and Digit have some form of autonomous banking in the form of autonomous saving, some companies want to take it even further. Envel, a not-yet-launched app running on Cambr, promises to provide what they call “Driverless Banking”, an AI powered banking framework to completely automate their customers’ financial lives.
Why Autonomous Finance can work
By providing the underlying infrastructure to fintechs spearheading autonomous finance, BaaS makes autonomous finance a reality.
- Start with investing: Until now, the most popular use case for autonomous finance has been autonomous investing. This is because the underlying infrastructure was available for autonomous investing. For autonomous finance to work, you need two things: A real time data stream and real time action potential. For example, with investing, you have hundreds of data streams, and using internal logic, a broker or asset manager can make personalized buy/sell decisions on behalf of an investor.
- Move to banking and personal finance: With banking, however, the underlying infrastructure has only started becoming available recently, due to Banking as a Service.
Challenges
The infrastructure autonomous banking needs in order to work and why older infrastructure doesn’t work for autonomous banking
In order for autonomous finance to work, the infrastructure needs to be able to take in real time data, handle large volumes of requests in real time, and have the ability to issue real time transaction instructions.
- Incumbents don’t (yet) have the infrastructure: The largest banks in the United States run on one of three legacy cores. These cores were built in the 1960s, and as the internet didn’t even exist, they did not contemplate the internet in their creation. Although these cores have been iterated many times over the years, they continue to be built on that underlying cobalt infrastructure, which is not able to process thousands of requests, real time data, or do anything actionable about it, since it wasn’t built for these tasks.
A quick example: To illustrate this, let’s take a case of a person who wants to limit his spending at restaurants to within a specific budget, and if he tries spending over that budget, his card would not work.If you tried to do that on a legacy core, you would likely overload the core.
This is because a few things are happening here:
- Intake real time data: Every time you get a new transaction, you need to understand the transaction and the merchant
- Ledger that against the customer: The customer has spent $20 at the merchant, and he now has another $25 until his budget is spent. This ledger needs to be maintained and the card needs to be locked if surpassed.
- Locking the card: In order for the card to be locked, the system needs to be able to issue a command to the card, in real time, that tells that card, “no more transactions whatsoever”. An even more sophisticated way this could work is by restricting spending on a specific category of transaction.If the merchant category is a restaurant, then lock the card. If the merchant category is not a restaurant, then allow the transactions.
So the way that flow would work is that when you go to the restaurant and try to purchase something, when you swipe your card, the processor sends a message to the core and lets the core know, “this is the merchant category”. Based on this logic, either allow or deny the transaction.
In order to facilitate that, the data needs to be processed in real time, decisioned in real time, and the card locked in real time. These legacy cores cannot do these things. So now, players like Cambr, Deserve, and Marqeta are making this possible and more available to firms that want to use autonomous finance as they consider how to service their customers.