PostsHome
EducationPersonal FinanceTechnology

The Hard Problem of Money

Pulse EditorialJanuary 12, 20269 min
The Hard Problem of Money
Illustration:The EconomistThe Economist

In March 2020, the S&P 500 dropped 34% in 23 trading days. Brokerage platforms reported record login activity between midnight and 4am. People who hadn't checked their portfolios in months suddenly couldn't stop refreshing.

The behavioral finance literature has a term for what happened next: panic selling. Nearly 10% of retail investors liquidated positions during the crash, according to research published in the journal Risks. Most did so at or near the bottom.

The market recovered in four months. The fastest rebound in 150 years.

Those who sold locked in losses. Those who stayed recovered everything and then some. The difference between the two groups wasn't knowledge or sophistication. It was whether someone talked them through the night.

The Replacement Thesis

Fahad Hassan, CEO of the Virginia-based wealth-tech startup Range, has been explicit about his intentions.

"We're trying to make that wealth advisor a computer," he told WealthManagement.com last year. "And that will happen. It's just a matter of when, not if."

Range raised $60 million in November 2025, bringing total funding past $100 million. The company manages $400 million and claims 300% year-over-year revenue growth. Its AI assistant, Rai, handles thousands of client queries monthly and has reduced messages to human advisors by 50%.

Origin Financial, meanwhile, launched what it calls "the first SEC-regulated AI financial advisor." In standardized CFP exam testing, Origin's AI scored 96.4% against an average of 79.5% for human advisors. A 17-point margin.

The pitch writes itself: better performance, lower cost, unlimited scale.

The pitch is also missing something fundamental.

What the Benchmarks Don't Measure

Vanguard has been studying advisor value for over two decades. Their framework, called Advisor's Alpha, quantifies the annual return premium that good financial advice provides.

The breakdown is instructive:

ServiceAnnual Value Added
Behavioral coachingUp to 1.5%
Asset allocation0.75%
Tax-loss harvesting0.75%
Rebalancing0.35%

The largest single component has nothing to do with portfolio construction or tax optimization. It's behavioral coaching: keeping clients from making catastrophic decisions during moments of fear or euphoria.

Vanguard's researchers found that emotional outcomes account for 45% of the total value clients perceive from their advisory relationship. The spreadsheet work accounts for the other 55%.

"AI doesn't feel anything and thus is incapable of empathy," Vanguard noted in a 2025 analysis. "While AI can simulate emotional responses, clients recognize the difference."

Two-thirds of Americans agree. A recent survey found 64% believe AI cannot understand how emotions affect financial planning.

The Casualties

The evidence of harm is no longer theoretical.

A Pearl.com study found that 20% of Americans have acted on financial advice from AI tools. Of those, 19% reported losing $100 or more. Among Gen Z users, the figure rises to 27%.

In the UK, the accounting software firm Dext surveyed 500 accountants and bookkeepers in December 2025. Half reported awareness of businesses that suffered direct financial losses from AI-generated tax and financial guidance. The damage included overpayments, missed allowances, penalties, and compliance failures.

"The damage is no longer hypothetical," said Paul Lodder, Dext's VP of accounting product strategy. "Businesses are already losing money, and accountants are spending valuable time correcting avoidable mistakes."

Nearly all accountants encountering AI errors reported spending additional time on corrections. Two-fifths estimated 4-10 hours per month undoing the damage.

In one documented case, an AI chatbot advised a British user to exceed ISA contribution limits. The recommendation, if followed, would have triggered HMRC penalties.

When AI advice goes wrong, there's no fiduciary duty, no malpractice claim, no one to call. The liability sits with the user.

The 3am Problem

Michael Kitces, one of the most widely read voices in financial planning, has written extensively about what he calls "the 3am problem."

It works like this: A client wakes up at 3am, unable to sleep, consumed by market anxiety. They open their brokerage app. The sell button glows.

In a traditional advisory relationship, many clients text or call their advisor. The advisor responds. Not with charts or data. Usually with something simple: I know this is scary. Don't do anything tonight. Let's talk tomorrow.

The intervention isn't analytical. It's relational. The advisor's job in that moment is to be a circuit breaker between fear and action.

AI systems are available 24/7. They can generate responses instantly. What they cannot do is recognize that the question being asked isn't really about money. It's about terror. About mortality. About the fear of losing everything.

An AI will answer the question as posed. A human advisor hears the question underneath.

Life at the Margins

Financial planning firms that specialize in life transitions report a consistent pattern: the moments when advice matters most are the moments when clients are least capable of processing it.

Divorce drops household income for women by 45% on average. The emotional stress mimics grief. Asset division decisions made in the first six months often prove suboptimal years later.

Widowhood is worse. Surviving spouses face Social Security elections that can swing retirement income by tens of thousands of dollars. Claiming at 60 versus 67 has permanent consequences. Most widows and widowers are in no condition to optimize anything.

Kathleen Marteney, a wealth planner in Ohio who specializes in this population, doesn't start with investment strategy. She starts by asking about the person who died. What were they like? What do you miss? How are you sleeping?

Then she waits. Sometimes for weeks.

"People in grief can't make good financial decisions," Marteney has said. "Not because they're incapable. Because they're grieving."

No AI startup is training models on the skill of strategic silence. No benchmark tests for knowing when to stop talking.

The Force Multiplier Argument

The most thoughtful voices in the industry aren't arguing for human advisors or AI advisors. They're arguing for both.

Mike Conover, CEO of the AI research platform Brightwave, put it directly: "You always want to have a human in the loop. We should make sure that humans are reviewing recommendations and that humans are the ones making final decisions."

Deloitte's Chief Futurist offered a sharper framing: "Too many business leaders see generative AI as merely a way to reduce costs by automating and eliminating jobs. You can't shrink your way to success."

The Business Agility Institute found that AI's value as a force multiplier peaks when it augments human decision-making rather than replacing it. Their research describes a model where people become "composers" who provide inputs, quality control, and creative judgment that AI cannot generate independently.

This isn't a compromise position. It's recognition of where the hard problems actually live.

Portfolio optimization is computationally intensive but conceptually straightforward. The math is knowable. The inputs are quantifiable. AI excels here.

But money is never really about money. It's about security, freedom, legacy. The fear of not being able to care for the people you love. These are human problems. They resist quantification.

The Persistence of Judgment

Range can build autonomous agents for compliance and tax optimization. Origin can score higher on CFP exams than the median human planner. Both achievements are real.

Neither addresses the core difficulty.

The hard problem of money isn't calculation. It's that people don't behave like spreadsheets. They panic at bottoms and chase at tops. They make emotional decisions about inheritances. They sabotage their own retirements out of fear or denial or magical thinking.

AI can calculate optimal withdrawal rates across 150 years of market history. It can backtest every allocation strategy ever conceived. It can tell someone, with decimal-point precision, exactly how much they need to retire at 65.

What it cannot do is notice that the question about retirement is really about a dying parent. Or that the request for "aggressive growth" masks a gambling problem. Or that the silence on the other end of the phone means the client needs to cry before they can think.

The best technology in the world can't replace the moment when someone says, "I hear you. This is hard. You don't have to decide today."

Where This Leaves Us

The question isn't whether AI will get smarter. It will.

The question is whether smarter is enough.

For the calculations: yes. For the conversations that actually change lives: the jury is still out.

What seems clear is that the path forward isn't replacement. It's integration. AI handling the analytical work. Humans handling the human work. Each doing what they do best.

The hard problem of money has always been the same. Not returns. Not allocation. Not tax efficiency.

The hard problem is that we're not rational about the things that matter most.

And that's not a bug to be patched.

It's just the truth about being human.

Sources:

Share this article
Back to all posts