
Introducing Autopilot
Today, we’re launching Autopilot: Cleo’s first step toward autonomous money management.
Closed

Financial advice has never been easier to find. It’s free, endless, and always a few taps away: Open any app, and someone’s ready to tell you how to budget, what to invest in, and which debt to clear first. If access to advice were what stood between people and their financial health, most people’s finances would be improving. Instead, money stress and financial setbacks remain near their highest levels in years.
What most people struggle with isn’t a shortage of advice. Worse, not all of that free advice is good. For every sound tip, there’s another that’s oversimplified, context-blind, or flatly wrong, delivered with just as much confidence. So the real question is why so little advice changes what people do, and why some of it makes things worse. The answer, I’ve found, is that changing how you handle money is more like building a muscle than following instructions.
Advice gives people knowledge, but knowledge is only one input into behavior. Hearing the right advice doesn’t mean someone will act on it, and advice that sounds right is sometimes wrong.
From the outside, a confident but inaccurate voice and a correct one sound identical. This is what makes confidence without real knowledge dangerous. Someone who acts quickly and decisively on bad advice can do more damage to their finances than someone who does nothing at all.
As a financial counselor, I’ve seen this play out in different ways. Take Priya and Marcus, both of whom paid attention to money content online:
Priya had done the reading. She could explain why a high-yield savings account beats idle cash and why low-cost index funds beat picking stocks. Yet her money sat untouched, because every time she went to open an investment account, she froze, sure that she was about to choose the wrong provider or start at the wrong moment.
Marcus, on the other hand, had a different feed. He saw confident threads about the stock about to run, a friend’s posts about his crypto win, and warnings that the cautious get left behind. In response to all this hype and urgency, he acted fast. By the time we spoke, he had moved a large share of his savings—including money he would need within the year—into a single volatile bet.
Priya knew what to do, but couldn’t find the confidence to act. Marcus had the confidence to start, but not the knowledge to choose well. Each was missing what the other had.
Getting someone to the point of acting takes two things: knowledge and confidence. The clearest way to see how they work together is to plot them against each other, with knowledge running low to high on one axis and confidence on the other.
The quadrants formed by those two axes reflect four very different people:
Low knowledge, low confidence: someone overwhelmed and unsure, prone to putting money decisions off indefinitely.
High knowledge, low confidence: someone who knows the right move but hesitates to act, so being right never turns into doing anything.
Low knowledge, high confidence: the risky actor (the most dangerous of the four), who acts readily on poor information and can do real harm precisely because the action feels so assured.
High knowledge, high confidence: someone informed and ready; the point where a good decision finally gets made.

The four quadrants of the knowledge–confidence matrix
This model explains something the “just give better advice” view can’t. Two people can hear the same tip and do completely different things with it, depending on which quadrant they occupy. The advice that nudges one person toward financial health can push another deeper into risky behavior.
The research backs this up. In a large US sample, Sajid et al. (2024) found that financial literacy and confidence both improve financial well-being, partly by improving day-to-day financial behavior. On the risky quadrant, Asaad (2015) found that confidence gets people to act at every knowledge level, but when it outruns real knowledge, the actions it produces turn risky and costly.
Similar patterns hold in other countries. In Italy, Aristei and Gallo (2021) showed that overconfident people invest more, but also lose more to fraud and carry more debt. In Canada, Morris et al. (2022) found that confidence plays a central role in explaining financial behavior. And in Brazil, Ramalho and Forte (2019) found that knowledge and confidence together predict better behavior, except among the over- and underconfident.
That top-right corner is just a starting point. Being informed and ready gets a good decision made once. It doesn’t yet make that decision a habit.
If people sit in different quadrants, the same advice can’t serve all of them. For some, more advice isn’t the right move at all. Good help starts with diagnosis, not prescription.
Before offering a single tip, the most useful question to ask is whether someone is short on knowledge, confidence, or both. A knowledge gap calls for education and clear, trustworthy information. A confidence gap calls for small wins, less shame, and encouragement that makes sound action feel possible.
Priya didn’t need another explanation of index funds; she could have written one. What changed things for her was opening an investment account in a single sitting and setting up one automatic transfer, so the decision got made once and the transfer kept running on its own. Confidence’s job is to take the knowledge you already have and turn it into a start, then momentum, then consistency over time.
Marcus needed the opposite. He had plenty of conviction, but was short on the knowledge needed to aim it accurately. He was acting out of fear of missing out rather than on any credible investing strategy, so our job was to slow him down. We separated the money he could not afford to lose, then walked through ways to invest the rest that balanced his timeline against his real risk tolerance. Along the way, he came to see that what he had been doing was closer to gambling than investing. More raw confidence was the last thing he needed.
Getting someone to start, though, isn’t the same as changing their behavior. Marcus is proof. He made the corrections and set up steady, automatic investing, and for a while it held. Then the market ran up, the confident posts came back, and within a couple of months he was eyeing another all-or-nothing move—the exact pattern we had just walked him out of.
A good decision is just one moment. Behavior change is what happens when that decision gets made again and again, until it no longer takes any deciding at all. It’s like a muscle: built through repetition and weakened the moment it goes unused. Knowledge tells you what the right movement is and confidence gets you to attempt it, but after the first rep, the work changes.
Instruction matters less than momentum. Priya’s single automatic transfer was one rep. What changed her finances was the fortieth, made without a second thought. Each consistent rep makes the action a little more automatic and hands back a little more confidence, which makes the next one easier.
Habits run in both directions, though. Sometimes the job is to interrupt a bad habit that already has momentum. That was Marcus, whose problem was never inaction. Instead, it was a confident habit pulling the wrong way, so consistency was the obstacle rather than the goal. Our work was to keep interrupting the old impulse until the disciplined choice became the default.
The job of a financial coach is bigger than giving more advice, or even better advice on its own. It’s to get someone to start, and then to help them stay consistent long enough that the behavior sticks. Confidence built on shaky knowledge only helps people act on the wrong things faster. But knowledge and confidence together get the first rep done, and consistency turns that rep into who they are. Give people that, and behavior change stops being something you have to push on them. It becomes something they own.
Robinson Torres is an Accredited Financial Counselor (AFC) and Cleo’s lead financial expert. He works directly with people to improve their financial health and well-being, and partners with Cleo’s product team to scale this impact to millions of users, helping turn sound financial guidance into lasting behavior change.
Aristei, D., & Gallo, M. (2021). Financial knowledge, confidence, and sustainable financial behavior. Sustainability, 13(19), 10926. doi:10.3390/su131910926
Asaad, C. T. (2015). Financial literacy and financial behavior: Assessing knowledge and confidence. Financial Services Review, 24(2), 101–117. https://openjournals.libs.uga.edu/fsr/article/view/3236
Morris, T., Maillet, S., & Koffi, V. (2022). Financial knowledge, financial confidence and learning capacity on financial behavior: A Canadian study. Cogent Social Sciences, 8(1), 1996919. doi:10.1080/23311886.2021.1996919
Ramalho, T. B., & Forte, D. (2019). Financial literacy in Brazil: Do knowledge and self-confidence relate with behavior? RAUSP Management Journal, 54(1), 77–95. doi:10.1108/RAUSP-04-2018-0008
Sajid, M., Mushtaq, R., Murtaza, G., Yahiaoui, D., & Pereira, V. (2024). Financial literacy, confidence and well-being: The mediating role of financial behavior. Journal of Business Research, 182, 114791. doi:10.1016/j.jbusres.2024.114791

Today, we’re launching Autopilot: Cleo’s first step toward autonomous money management.

When extending Cleo’s chat engine to real-time voice, we needed to keep Cleo’s personality and tone while maintaining high accuracy and low latency.

Cleo’s quick replies only help if they arrive before users start typing, so we fine-tuned a specialized small model to reduce latency.

It’s part of a larger system that keeps message classification accurate as user behavior shifts and new agents come online.