By Lynn Räbsamen, CFA | COO, Global Swiss Learning | Advisory Board Member, CFA Institute | Author, Artificial Stupelligence
A $60 Seat, Billions in Lost Value
In February, a fintech called Altruist switched on an AI tool named Hazel. It reads a client’s tax return, pay stubs, account statements and meeting notes, then drafts a personalized tax strategy in minutes — for $60 per advisor seat a month.
The market did not find this charming.
Within days, wealth-management stocks sold off. Raymond James had its worst session since the 2020 crash. Over the following week, Schwab, LPL and Ameriprise each slid by double digits. Bloomberg Intelligence summed up the fear in three words investors understand: efficiencies competed away.
By June, Bloomberg was describing advisors who earn upwards of $500,000 as facing a “chatbot reckoning.”
The market’s read was simple: if a machine can do the work, the fees come down. That read is half right. The other half is where the real risk is hiding.
“The market priced fee compression correctly. It didn’t price the real risk.”
What the Selloff Got Right
Fee compression is real, and it is not a forecast. It is arithmetic.
The traditional advisor’s economic moat was rarely the investment insight. It was the labor — the tax prep, the account paperwork, the document-wrangling that justified a percentage of assets every year. Hazel automates exactly that. Altruist’s CEO has cheerfully predicted that within a year, most of an advisor’s existing tech stack will be “irrelevant.”
When the moat was administrative, and the administration is now a feature, the fee has nowhere to hide.
“When the moat was paperwork, automation doesn’t threaten the advisor. It repossesses the margin.”
So far, the market and I agree. The advisor who charged for busywork should be nervous. But notice what this version of the story quietly assumes: that the client is migrating to a better, cheaper tool.
That is not what the client is doing.
The Risk No One Put a Number On
Clients are not defecting to Hazel. Hazel is sold to advisors. Clients are defecting to the free chatbot in their pocket.
A Credit Karma survey found 66% of respondents had used generative AI for financial advice — rising to 82% among millennials and Gen Z. They did not compare providers. They opened an app and asked.
Here is the figure nobody discounted into the share price. A widely cited 2025 study by Investing in the Web put 100 finance questions to ChatGPT and had experts grade the answers. Roughly a third were unsuitable: 29% incomplete or misleading, 6% flatly wrong. MIT’s Andrew Lo has been blunter still, naming finance one of three domains — alongside law and medicine — where bad AI advice does real damage.
So the threat to the advice industry is not a machine that advises better. It is a machine that advises confidently, for free, and incorrectly about a third of the time — with no suitability assessment, no fiduciary duty, no traceable rationale, and no file a regulator could ever review.
“The competition is not better advice. It is worse advice that costs nothing and never asks whether you can afford the risk.”
Cheaper Is Not the Same as Better
The uncomfortable mechanism underneath all of this is that financial advice is a market where the buyer often cannot judge the product.
The people leaving are largely the ones the industry priced out anyway. Schroders found three in four advisers won’t take a client with under £50,000 to invest. For them, advice was never a relationship. It was a fee they couldn’t reach. A confident chatbot doesn’t have to be good. It only has to be free and available, in a market where “good” is invisible until something breaks.
That is how fees compress without quality improving. Not because AI won the argument, but because the client stopped being able to hear it.
“Fee compression isn’t the market rewarding better advice. It’s the market forgetting how to tell the difference.”
A Hidden Opportunity — With a Catch
There is, theoretically, an opportunity buried in all of this. At $60 per advisor seat, Hazel could change the unit economics of serving a £50,000 client. The administrative cost collapses; the relationship becomes viable. For the first time, the mass-affluent segment — the one currently turning to ChatGPT by default — could be served by a real advisor operating inside an accountable, regulated framework.
The catch is that no advisor earning $500,000 a year has historically felt the urgency. Downmarket is not a direction high earners typically volunteer to go.
But if AI continues to compress fees at the top end, that calculus changes. The advisor who once wouldn’t return a call from a £50,000 client may find that serving the underserved is no longer a charitable impulse — it is the only business model left. Or they find another line of work entirely.
The irony would be complete: the technology that threatened the most lucrative end of the market ends up redirecting talent toward the end that needed it most.
A Closing Observation
Wall Street sold off the advisors because a machine learned to do the paperwork. Fair enough. But the paperwork was never the danger.
The danger is the advice clients are quietly taking instead — fluent, instant, unaccountable, and wrong often enough to matter — filed under “savings” until the day it isn’t.
The market is very good at pricing the risk it can see. This one it mistook for a discount.
For more insights about what AI can or cannot do, check out my book “Artificial Stupelligence: The Hilarious Truth About AI“.
Subscribe here to be the first to receive my insights.
