Gallup and Edward Jones published a study on August 5 that deserved more attention than it got. Three-quarters of American adults went looking for financial guidance in the past year. Seventy-three percent did their own internet research. Thirty-five percent asked a family member. Thirty-two percent talked to a professional advisor. Twenty-six percent got it from news, media, or social media. Nineteen percent asked an AI tool.
Then Gallup asked how much confidence people have in each source. Financial advisors, seventy-nine percent. Finance professors, sixty-four. Your relatives, sixty-two. Your friends and coworkers, fifty-four. Artificial intelligence, twenty-seven. Social media influencers, thirteen.
Only three percent of Americans have a great deal of confidence in AI for money questions. Forty-five percent have none at all. I have spent the last two years building production agent systems — MCP servers, retrieval pipelines, the observability layer that catches these things when they go sideways — and my honest reaction to that number was: yeah, that sounds about right.
But I got curious about the twenty-six percent taking their guidance from news, media, or social media, because that bucket is where the famous people live. So I ran a small experiment. I picked seven questions ordinary people actually ask, went and found what four of the most-quoted money figures in America have verifiably said about each one, then put the same seven questions to a frontier model in a clean session — no system prompt, no follow-ups, the way somebody would at eleven at night.
Everything attributed to a named person below is documented and sourced. No impressions, no invented quotes. That constraint mattered more than I expected, because the real answers are stranger than anything I would have written for them.
The panel: Jim Cramer, who hosts the stock-picking show. Dave Ramsey, who owns the debt-and-radio franchise. Suze Orman, the most quoted retirement voice in America. And Warren Buffett, included as a control, because he is the one man on the list who unambiguously beat the market for decades.
Seven Questions
1. Should I pay off my credit card or invest?
Cramer: Pay the card, and it is not close. He has built entire episodes on this, going back to Stay Mad for Life. His framing is that even a one-in-a-million investor cannot outrun a balance compounding at credit card rates. He also puts two things ahead of investing that almost nobody else on television bothers to mention: health insurance and disability insurance. He is blunt about it, he is right, and the medical bankruptcy numbers back him up.
Ramsey: Pay the debt, and pause the 401(k) while you do it — including forgoing the employer match. He frames it as temporary, roughly eighteen months, and has said he would never tell anyone to stop for several years. He also does not hide the tradeoff. He has said outright that he knows a hundred percent match is a good deal, and that in his experience personal finance is eighty percent behavior and twenty percent head knowledge. He is choosing the behavior on purpose.
The machine: Capture the full employer match first, then throw everything else at the card. Its reasoning was that a match is an instant fifty to one hundred percent return before the market does anything at all, larger than any credit card rate you are likely to be carrying.
My read: The machine has the better of it, but not by as much as the internet believes, and Ramsey deserves more credit here than he usually gets. Run it properly and the match still wins — the contribution goes in pre-tax, which roughly offsets the tax you eventually pay coming out, and you finish ahead over a two-year payoff even against a twenty-two percent card. But the margin is thinner than “free money” implies, it narrows fast at penalty rates near thirty percent, and it inverts entirely if your match sits behind a vesting cliff you will not reach. What actually bothers me is not the advice. It is what happens to the advice in transmission. Ramsey states plainly that he is making a behavioral argument and losing the math deliberately. By the time it reaches a forum or a comment thread it has become a claim about the math. The message degrades on the way down, which is something I watch happen to technical guidance every single week.
2. Should I pick individual stocks or just buy an index fund?
Cramer: This is the answer that broke my first draft. The man who hosts the stock-picking show tells you to index if you will not do the work. His stated threshold is about four hours a year per stock for a casual observer, an hour a week per position for anyone serious, and no more than five names. Below that bar, he says, you default to an index fund. His recent book puts roughly half the portfolio in index funds with the rest across a handful of fast growers, and he has called a cheap S&P 500 fund the least bad way to manage money passively.
Buffett: The instruction in his 2013 shareholder letter, written for the trust that will hold his own widow’s money, was ninety percent into a very low-cost S&P 500 index fund and ten percent into short-term government bonds. He named the provider he had in mind. Not Berkshire. Not a hand-picked value portfolio. Not one of the many managers who trained under him. An index fund. He also settled the argument with cash: his 2007 million-dollar wager against a fund-of-funds ran ten years, and the hedge fund side conceded before the clock ran out.
The machine: Essentially Buffett’s answer, with a caveat I thought was sharper than expected — that ninety-ten was designed for a specific elderly beneficiary inheriting an amount large enough that the allocation barely mattered, and a normal person should size equity to a drawdown they can actually sleep through.
My read: Here is the part that stuck with me. In March 2020, mid-crash, Cramer told viewers to stay away from index funds because owning everything meant owning the losers too, and to buy recession-proof consumer names instead. Same man. Different week. I do not read that as hypocrisy. I read it as what a nightly broadcast does to a person who has to say something new at six o’clock. The book gets his considered view. The show gets his Tuesday. Anyone who has shipped software under a release cadence they did not choose knows exactly how that happens.
3. How much can I safely withdraw in retirement?
Ramsey: Eight percent. His reasoning is arithmetic: stocks return about twelve percent, inflation runs about four, so eight comes out of the growth and the principal never gets touched. He has dismissed the widely used four percent guideline as flatly wrong and, in his phrasing, ridiculous. He has told callers with seven-figure balances they can pull eight percent indefinitely.
Orman: Three percent.
The research: Morningstar’s December 2025 retirement income work, which uses forward-looking return assumptions rather than backward-looking averages, put the highest safe starting withdrawal rate for a new retiree seeking a ninety percent chance of not outliving the money over thirty years at 3.9 percent.
The machine: Roughly four percent as a starting point, and it went straight to sequence-of-returns risk without being asked — that a retiree does not experience an average return, she experiences one particular sequence in one particular order while pulling money out along the way, and a bad sequence early is what empties the account.
My read: Ramsey’s arithmetic is correct and his model is wrong, which is the most dangerous combination available in any discipline. Twelve minus four does equal eight. But an average annual return cannot express path dependency, and path dependency is the entire reason the four percent literature exists. I have spent a career watching people mistake a working average for a working system, and this is that failure mode with someone’s retirement inside it. Independent testing of an eight percent rate finds it survives a minority of historical thirty-year retirements — the estimates I found ranged from one in five to about a third, which is itself worth noticing, because the people checking his number cannot agree on how badly it fails.
Also worth saying plainly: This is the widest disagreement on the page, and it is not between the humans and the machine. Ramsey says eight, Orman says three. They are shelved next to each other at the airport.
4. How much do I need to retire early?
Orman: Five million dollars minimum, and possibly ten. She has said two million is nothing. Her arithmetic, which she has explained openly, works backward from what it cost to care for her mother in old age — roughly thirty thousand dollars a month — plus living expenses, landing near three hundred fifty thousand a year after tax. At a five percent draw that requires seven figures several times over.
The machine: It gave a number, then immediately asked what it arguably should have asked first — annual spending, healthcare exposure before Medicare eligibility, whether the house is paid off. Fed a household spending eighty thousand a year, it worked backward from a withdrawal rate rather than forward from a headline: about two million at four percent, closer to two and a half at the lower rate a forty-year horizon deserves.
My read: Orman generalized from one person’s life, and the person is a multimillionaire broadcaster reasoning from a private long-term-care bill most families will never see at that price. The underlying warning is sound — a forty or fifty year horizon means a catastrophe you cannot work your way through — but the number is not a number. Be honest about the sensitivity, too: load in twenty-five thousand a year of pre-Medicare healthcare for a family and that eighty-thousand household needs something closer to three and a quarter million. That is a real adjustment. It is still not five, and it is nowhere near ten. And notice what actually separated the two answers here. It was not intelligence. She has to give one figure to a national audience and the machine gets to ask three questions first. That is a structural advantage, and it is most of what people are responding to when they say the chatbot understood them.
5. The market just dropped. Should I sell?
Cramer: Rule 7 of his twenty-five rules is that nobody ever made a dime by panicking. He has been saying it for twenty years. He goes further: he has advised investing twice your normal amount whenever the market falls ten percent or more, and doubling again the next month if it stays down.
The machine: No, and it made the point that investors who do worst are reliably the ones who convert a paper loss into a realized one at maximum pessimism, then sit out the recovery waiting to feel confident again — confidence being a lagging indicator. Its structural suggestion was that if the drawdown made you want to sell, the allocation is more aggressive than your stomach, and that gets fixed by rebalancing when things are calm.
My read: They agree on the first half and I agree with both. Where I part company is the double-down-on-dips rule, which is market timing wearing a cardigan. It assumes you are holding idle cash for precisely this purpose, which most people are not, and it quietly requires you to be braver than average at the exact moment being average is the winning move. Consistent contributions on a schedule capture most of the same benefit without asking you to correctly identify a bottom, which nobody does twice.
6. How much should I hold in my own company’s stock?
Cramer: Never more than twenty percent of your savings in any single stock, and he calls diversification the only free lunch in the market. He specifically flags over-concentration in your employer as one of the most common mistakes he sees.
The machine: Agreed with the twenty percent ceiling and went lower for employer stock on correlation grounds.
My read: Go lower than both of them. I was employee number six at a startup, and I have watched enough equity concentration up close to have opinions about it. The danger is not that the position is large. It is that the position and the paycheck fail together, in the same week, at the same moment the local job market for your exact skill set is at its worst. Three correlated losses arriving simultaneously is not a diversification problem, it is a single point of failure, and every engineer reading this already knows what we call a system with one of those. Enron was never a diversification story.
7. Should I own gold or crypto?
Cramer: Five to ten percent in what he calls a hedge, and his two named favorites are gold and bitcoin. His framing is insurance rather than growth — something that might behave independently of the stock market and hold value if things go badly wrong elsewhere. He is explicit that he does not expect it to make you rich.
The machine: It declined to take a side on the assets and said the framing was what mattered — an allocation held as insurance has a defined job, a size limit, and a rebalancing rule, while the same allocation held because you expect it to go up is a directional bet.
My read: That distinction is the whole ballgame and almost nobody honors it. You can tell which version somebody is actually running by whether they check the price. Insurance you do not check. Cramer’s framing is more defensible than either side of the gold argument, and it is the sort of thing that gets lost when the segment needs a number and a graphic.
Now the scoreboard
The joke only pays if the record is real, so I went and checked it.
Cramer has been measured repeatedly by people with tenure. The most complete study is Hartley and Olson in The Journal of Retirement, 2018, tracking his Action Alerts PLUS portfolio from 2001 through 2017 — an honest window, because it includes the post-crisis years earlier and friendlier studies stopped short of. Annualized return over that span: 4.08 percent. The S&P 500 over the same period: 7.07 percent. Underperformance on raw return and on Sharpe ratio, with factor analysis attributing most of the result to underleveraged market exposure and tilts toward small caps, growth, and poor earnings quality. Which is a polite academic way of saying the alpha was beta in a costume.
So bet against him. People tried. In March 2023, Tuttle Capital launched two funds for exactly this purpose, one tracking his calls and one inverting them. The long fund closed after six months. The inverse fund closed five months later, down roughly fifteen percent since launch, while the S&P 500 gained twenty-five percent over the same stretch. It had gathered about two million dollars. Following him lost to the index. Inverting him lost by more. You cannot profitably invert noise; you just pay the spread twice.
Ramsey has been measured too. The Money Guy team ran his recommended four-fund allocation — growth, aggressive growth, growth and income, international — across trailing periods. Over twenty years it annualized around 9.8 percent. Over the ten years from 2013, around 10.8 percent, against an S&P that did better than fourteen. Those are respectable returns. They are also not twelve percent, and twelve percent is the number the entire eight percent withdrawal rule rests on. His own recommended portfolio does not produce the input his own retirement math requires.
What actually surprised me
I started this expecting the comedy to live in the distance between a chatbot and a man with a soundboard. That is not where it is.
Line the answers up and the humans are not one position. They are four positions, and several of them are violently incompatible. Ramsey says withdraw eight percent; Orman says three. Ramsey says skip the match to kill debt; Cramer wants the debt gone but is not asking anyone to walk past free money. Orman says you need five million to stop working; Cramer says a casual investor can get there on four hours a year and an index fund. And Buffett, the only one of the four with a documented multi-decade record of beating the market, instructs his own widow not to try.
On six of seven questions the machine landed closer to Cramer than Cramer landed to Ramsey. That was not the article I set out to write. The man who is a national punchline for stock picking turns out to give conservative, structural, almost dull advice — pay the card, get insured, cap any single position, index if you will not do homework, do not panic — and the reason nobody knows that is that the structural advice goes in the book, which sells once, while the urgency goes on the air, which sells every weeknight at six.
That is not a Cramer problem. It is a format problem, and I recognize it because it is the same one I write about in enterprise software every month. The demo is exciting. The thing that actually ships is a queue, a retry policy, and a dashboard nobody wants to look at. Nobody ever built a nightly audience out of diversify and wait.
Where the machine falls down
It would be cheap to spend three thousand words on other people’s records without turning the lens on the tooling I build for a living.
The strongest research here is a paper by Taha Choukhmane, Weidong Lin, and Matthew Akuzawa at MIT Sloan with Tim de Silva at Stanford, which won the Swiss Finance Institute Outstanding Paper Award this year. They built a life-cycle simulation, asked a thousand Americans to write their own prompts to an AI financial advisor, and modeled what happened if those people followed the answers for decades.
The headline is genuinely good. AI advice on average nudged people toward sound long-term behavior — de Silva’s framing was that it is not perfect but it beats how many people currently decide, which is asking friends and family or running a quick search. Given that Gallup has seventy-three percent of guidance-seekers doing their own internet research and thirty-five percent asking relatives, that is the bar that actually matters.
The catch is serious. Advice quality tracked prompt quality. People with lower financial literacy wrote prompts carrying less information about their own situation and got back thinner, less personalized guidance. Prompts written by men and by more financially literate users drew higher recommended equity allocations. Over a simulated lifetime that compounded to roughly five percent more wealth near retirement — on the order of fifty thousand dollars by age sixty. Choukhmane put it plainly: regular people do not write prompts the way a finance professor does.
Prompt quality is now a form of financial literacy. Which means the machine does not close the advice gap. It re-encodes it in a friendlier font. And I want to name the failure mode I see every single day at work, because it is the one that matters here: these systems have no priors to defend. Push back and they fold. Every one of the four humans above would have held their ground, and on some questions that stubbornness is the entire value of asking a person. A model that agrees with you is not giving you advice, it is giving you a mirror with better grammar.
I spend my days building retrieval layers whose whole job is stopping a model from answering confidently from memory when it should be answering from a source of truth. Ask one about your retirement and there is no source of truth in the loop. There is just the model, being agreeable, in a very nice font. The operational failures are already showing up in the wild — wealth managers told CNBC last week about a model hallucinating details of a client’s trust document, about being told two ETFs were identical when one was equal-weighted and the other cap-weighted, about capital gains math that came back simply wrong. Every one of those reads as authoritative right up until it costs somebody money.
The number that should actually bother you
Back to Gallup, to the cut nobody put in a headline.
Among adults the study classifies as financially fulfilled, sixty percent work with a professional financial advisor. Among the financially stressed, fourteen percent do. AI use runs the other direction — twenty-one percent among the stressed, fifteen percent among the fulfilled.
The people with the most money get a human being with a fiduciary duty, a license that can be revoked, and a phone number that rings. The people with the least get a chatbot and a broadcast. Both are free at the point of use. Neither can be sued by you.
That is not an AI story. It is the same two-tier structure I have written about in retirement and in healthcare and in legal services — the expensive tier comes with accountability, the free tier comes with volume. AI did not create the split. AI made the bottom tier fluent enough that it is genuinely hard to tell you are standing in it.
Which brings me back to twenty-seven percent. I do not think that number is a failure of public understanding, and I do not think it should move much. But the skepticism deserves to point in every direction it applies. A man with a nightly slot is not a fiduciary. Neither is a radio host who has repeated the same wrong withdrawal rate for two decades without a single consequence, because nobody in this business gets fired for being wrong about your retirement. They get fired for being boring.
Use the machine to sharpen the question. Use the humans for the parts they are documented to be good at, which is more than the internet gives them credit for. Then go find somebody who has to answer for the answer.
Where this came from
Survey data from “Where Americans and Canadians Turn for Financial Guidance,” Gallup and Edward Jones, August 5, 2026, based on web surveys of 5,075 U.S. adults conducted March 20 to April 6, 2026, margin of error plus or minus 1.8 points. Cramer’s positions are drawn from CNBC and TheStreet reporting on Mad Money and from his books Real Money, Stay Mad for Life, and How To Make Money In Any Market. Ramsey’s withdrawal and return assumptions are documented in TheStreet, Yahoo Finance, and Money Guy’s portfolio analysis; the Morningstar 3.9 percent figure is from its December 2025 retirement income research. Orman’s figures are from her appearance on the Afford Anything podcast. Buffett’s allocation instruction is from the 2013 Berkshire Hathaway shareholder letter. Performance data from Hartley and Olson, The Journal of Retirement, 2018, and from ETF.com and ETF Stream on the Tuttle fund closures. AI research from “AI Financial Advice: Supply, Demand, and Life Cycle Implications,” Choukhmane, Lin, Akuzawa, and de Silva, via MIT Sloan and Stanford Report. Practitioner failure modes from CNBC, July 30, 2026.
None of this is financial advice. I am an architect, not an advisor, and the point of the piece is that you should be suspicious of confident-sounding money opinions from people with no obligation to you. That includes mine.
About the Author
Gal Ratner is the founder and CTO of Inverted Software and WhiteStar Labs, and Chief Architect at Prana Entertainment. He has spent nearly thirty years shipping production software on the Microsoft and .NET stack for clients including Microsoft, Sony, Rockstar Games, 2K Games, Best Buy, and Allegiant Air. He was employee number six at Break.com and a finalist for the Los Angeles Business Journal’s CTO of the Year.
His current work is in production agentic AI — MCP servers, the Microsoft Agent Framework, RAG pipelines, SQL Server 2025 vector search, and the PLogger observability framework — which mostly means he spends his days finding out where confident-sounding machine output meets a database that disagrees. He also operates ShopSnap, a zero-transaction-fee e-commerce platform built for agent-native commerce.
He writes as The .NET AI Guy That Ships, co-hosts Edge Grip Podcast, trains Brazilian jiu-jitsu under Sergio Penha in Las Vegas, and is the author of the novel The Archive of Lost Suns. He has never beaten the S&P 500 and has stopped trying.