Eleanor changed firms after she lost Tom. An algorithm met her first.

The short version:

More and more financial firms now use AI to score new clients and recommend portfolios, and those models are trained on a past that gave women less money and shorter credit histories. That can quietly produce more conservative, less tailored advice for women, often at the most vulnerable moment, just after a spouse’s death or an inheritance. You do not have to understand the math to protect yourself. You have to ask one question, and refuse to be talked out of it. Here is how it works.

Eleanor called us in March, about four months after she lost Tom. She had done the thing that a lot of women do in that first year, the thing that surprised even her, which was to move her accounts to a new firm. (Estimates of how often this happens vary quite a bit, from around twenty percent in Cerulli’s recent data to as high as seventy percent in older industry figures, so let’s just say Eleanor was in good and crowded company.) She had filled out the new firm’s online questionnaire one evening at the kitchen table, the laptop glowing, a cup of tea going cold beside it, and a few days later a portfolio recommendation arrived in her inbox with her name at the top and the word “Conservative” sitting quietly in a little gray box near the corner.

She was not sure she agreed with the little gray box. She wanted to know how it had decided that about her.

It is a better question than it sounds, and the honest answer is that, increasingly, no one at the firm can fully tell her.

The stranger who has read about you

Picture the system on the other side of that questionnaire as a very well-read stranger. It has studied the financial records of millions of people. It has noticed every pattern in how they saved, earned, paused, and spent. And then it meets Eleanor for the first time as a row of answers on a form, and it sorts her by resemblance to everyone it has seen before.

Here is the trouble with being sorted by resemblance. The records that stranger studied were written in a world that paid women less, interrupted their careers more, and titled the household accounts in the husband’s name. (The retirement numbers still carry the fingerprints of that world. Women reach retirement with about thirty percent less saved than men, and only twenty-two percent of women have a hundred thousand dollars or more put away, against thirty percent of men, per Census data summarized by Morgan Stanley.) The model is not being unkind to Eleanor. It is doing exactly what it was built to do, which is to repeat the past with great confidence. The past just happens to be a place where a woman’s thinner account history was read as a smaller appetite for the future.

You can think of it as inherited judgment, a kind of sediment that settles at the bottom of decades of data and gets stirred back up the moment Eleanor hits “submit.”

It does not need to know you are a woman

The part that surprises people most is that none of this requires the system to know Eleanor’s gender at all. You can leave the box off the form entirely, and the patterns still find their way in through the side door, through zip code, through the shape of a career, through who was listed first on the old joint account.

The clearest illustration the public ever got was the Apple Card episode a few years back, when a well-known entrepreneur reported that he had been offered a credit limit twenty times higher than his wife’s, even though her score was better. The card’s issuer did not, in any direct sense, ask about sex. The disparity arrived anyway. And in lending research published out of the University of Illinois, women have been shown to receive credit scores several points lower than men even after you account, carefully, for actual risk. A few points does not sound like much. On the wrong side of a threshold, a few points decides things.

Why this lands harder as we age

Two facts make this more than an abstraction for the women we work with.

The first is longevity. Women live about five years longer than men on average, and roughly four out of five people over eighty-five are women (Morgan Stanley, drawing on federal life tables). A small misjudgment about a thirty-year retirement is not a small thing. It compounds, quietly, for every one of those extra years.

The second is the inheritance wave that is arriving right now. Of the roughly fifty-four trillion dollars passing between spouses in the coming years, about ninety-five percent is expected to go to women, most of them as widows (CNBC). Which means an enormous number of women are going to do exactly what Eleanor did, walk into a new firm at the most tender moment of their financial lives, and be met first by a model rather than a person. The handoff and the algorithm arrive on the same afternoon.

The room where the rules have not been drawn yet

Here is what we find genuinely fascinating, in the way a slightly maddening puzzle is fascinating. We are trying to regulate a thing whose edges nobody has drawn.

In the United States, there is still no rule written specifically for artificial intelligence in investment advice. Regulators are stretching a law from 1940 to cover it, a law written for a human advisor who could be asked, out loud, why. That worked when there was a “who” to ask. One 2025 analysis, as reported in the trade press, estimated that the large majority of registered robo-advisors run on models that cannot actually explain their own decisions. So when Eleanor asks how the gray box decided “Conservative,” she has stumbled onto the exact question the rules were never built to answer.

The usual reassurance is that a human reviews the output. And often one does. But research on automation keeps finding the same human habit, which is that people defer to the confident machine even when it is wrong, in some studies the better part of nine times in ten. (We are all a little guilty of this. It is why we follow the map app down a closed road.) “A human reviewed it” can quietly come to mean “a human clicked approve.”

Europe is about to test a different answer. On August 2 of this year, the EU’s new AI law begins binding the high-risk uses, and it names credit scoring and the pricing of life and health insurance among them. It demands logs, real human oversight, and an explanation a person can follow. The United States has nothing like it at the federal level, and the states are only beginning to move. Colorado, for one, has a narrower law (SB 26-189) that takes effect in January 2027 and gives people the right to an understandable explanation of a decision made about them, a way to correct bad data, and a route to a human review. For now, though, a woman in Lyon has more recourse against an insurance algorithm than a woman in Louisville.

What’s worth doing

You do not need to become a computer scientist to hold your ground here. You need to do one human thing, which is to ask the question and refuse to be talked out of it. When a recommendation arrives with a label on it, ask the firm to explain, in plain language, what put it there and what would change it. A good advisor can answer that without flinching. If the only available answer is some version of “the system decided,” that is not a fact about you. That is a fact about the system, and it is worth knowing before you sign.

(Eleanor, for what it’s worth, was not “Conservative.” She was careful, which is a different thing, and her plan ended up looking nothing like the gray box. But that is a conversation for another afternoon.)

Every story is different, and every plan worth its name reflects that. If you would like to talk through what is going on in your world, our strategy session is forty-five minutes, virtual, and meant for listening, not selling. We use the time to ask the questions worth asking, to hear what is on your mind, and to understand what you are hoping a plan can do for you. Nothing is decided in the room. It is a conversation, and a careful one.

Disclosures:

Names and specific figures in client examples are illustrative. Specific outcomes depend on your individual circumstances. Statistics are drawn from the sources listed above; where a figure is a projection or industry estimate, it is described as such. Rose Line Retirement is a women-led retirement planning practice operating under Theus Wealth Advisors, an independently owned and operated Registered Investment Adviser. The content here is general education, not personal advice. For guidance specific to your situation, please speak with a qualified financial advisor, CPA, or attorney.