Financial Automation: How Far Should It Go?

More and more apps and banks promise to take the work out of managing money, by saving automatically, investing without a second thought, or paying bills without any intervention. The appeal is real: less friction, fewer mistakes caused by distraction, less temptation to spend what should have been saved. But how far should this automation go before it starts to replace, rather than support, the financial judgment of the person using it?

The short answer is that automation should be the hands, never the brain. It should carry out decisions we have already made - not make them for us.

The levels of financial automation

Financial automation isn't a single category. There are tiers.

At the most basic level are automatic transfers to savings or investments on payday. A little more sophisticated is the round-up, where each purchase is rounded up and the difference goes into savings. More advanced still are robo-advisors: platforms that invest the user's money and adjust the portfolio over time, without the user having to decide on each move manually.

A concrete example of what well-designed automation can achieve is the U.S. app Digit, which analyses a user's balance and spending patterns and moves small amounts into savings without them feeling the difference. A study by the Financial Health Network, published in 2022, examined Digit usage data and found that, for each month of active use, the saved balance increased by an average of 217 dollars — largely because saving decisions no longer required conscious effort every single time. The company itself has publicly reported that, just a few years after launch, its users had already saved more than a billion dollars in total through these automatic savings mechanisms, with substantially higher accumulated figures in more recent years.

When automation works well

Automation works well when it replaces an action that, on its own, is too small to justify a conscious decision each time, but that has a real impact when repeated. No one is going to stop and decide whether to save the 40 cents of change from buying a coffee. Automated, that micro-decision requires no effort at all and accumulates on its own.

This is the case with Acorns, an app that rounds each user purchase up to the nearest dollar and automatically invests the difference through its Round-Ups feature. Over the life of the app, its customers have collectively invested billions of dollars based on small everyday round-ups, and recent data from the company itself points to an average of around 45 dollars invested per month, per user. The underlying decision, "I want to invest my spare change", was made once, at sign-up. From then on, automation simply executes it, purchase after purchase, without asking for confirmation.

The risk of silent failures

The most subtle problem with financial automation is failing without warning anyone, that is, continuing to appear to work just fine while, underneath, it is harming the user.

The clearest case is that of the robo-advisor Betterment. In 2023, the SEC (the U.S. securities markets regulator) sanctioned the company over failures in its automated "tax-loss harvesting" service, a technique that sells positions at a loss to reduce the tax an investor pays on gains. Between 2016 and 2019, an undisclosed change in how often the software checked accounts, a programming limitation, and several coding errors prevented the system from applying this technique in more than 25,000 accounts, with nothing flagging the failure to the affected clients. The accounts kept operating normally, but the expected tax benefits were not being generated. In total, clients lost roughly 4 million dollars in potential tax benefits, and Betterment agreed to pay 9 million dollars in penalties and compensation.

This is the central risk of any automated system: it has no instinct for recognising when it is making a mistake. Someone who manages their finances manually quickly notices if something looks off. An automated system only stops when someone realises the error and that can take time.

Automation and financial literacy

There is a second cost, one that gets talked about less: the more we delegate, the less we learn. If you never have to decide how much to save or how to invest, you never develop the intuition to make those decisions when automation doesn't cover the situation. The psychologist Barry Schwartz described the so-called "paradox of choice": having fewer decisions to make reduces mental fatigue and can heighten a sense of relief — which helps explain the upside of automation. But the same mechanism that relieves can also distance a person from their own money, if the concrete experience of deciding disappears almost entirely.

What automation should not do: recognising when the rule no longer serves

Automation is good at repeating rules. It is bad at recognising when circumstances have changed and the rule no longer makes sense.

A well-known example of this is the so-called "Flash Crash" of May 2010. In about half an hour, the Dow Jones (one of the main indicators of the U.S. stock market) fell nearly 1,000 points, more than 9% of its value and recovered most of that drop in the minutes that followed. The fall wasn't triggered by fresh economic news at that moment, but by a large investor who used an automated system to rapidly sell a high volume of futures contracts into an already nervous market. Other automated systems, programmed to react to market patterns and volume, interpreted that sale as an alarm signal and began selling too, which generated still more automatic selling a spiral that none of the systems involved had been designed to recognise as anomalous. The situation was only resolved when human operators stepped in manually and temporary trading halts were applied.

None of the algorithms involved was technically broken. Each one correctly followed its own rule. The problem was that none had any way of perceiving that the collective context had stopped matching what its rules assumed and none paused to ask whether they still applied.

Where to draw the line

The most useful distinction isn't between "automate" and "don't automate," but between automating execution and automating the decision.

It makes sense to automate recurring transfers to savings, periodic investment in products already chosen consciously, and payment of interest-bearing debt. These are actions a person has already decided, once, that they want to repeat every time.

It does not, however, make sense to fully automate decisions about large sums or changes in investment strategy. The same applies to tax optimisation, which has legal nuances and calls for judgment about a context that keeps changing. It is an area where the examples of silent failures clearly show the risk of trusting automated processes blindly.

In practice, this means letting automation handle the day-to-day, the "base system", and reserving periodic manual reviews, quarterly or annual, to reassess whether the automatic rules still make sense. Automation executes. The person keeps deciding.

Conclusion

Financial automation isn't an all-or-nothing choice. It is a tool for making it easier to follow through on decisions already made. Used well, it eliminates the small everyday slips like forgetting to save or giving in to the temptation to spend. Used to excess, it creates the illusion that everything is under control precisely at the moments when human attention is most needed: when a new situation arises, or a silent error that no one has caught. The hands can, and should, be automatic. The decisions remain ours.