I Like AI. But I Worry About the Midnight Questions.
Artificial intelligence is a remarkable tool, and I say that as someone who uses it. I've watched it draft, summarize, translate, explain things faster than I can look them up, and take work that used to eat an afternoon down to about ten minutes. Nearly everyone I know is using it for something now — recipes, travel plans, homework help, work they'd otherwise be up late finishing.
We've all seen the good stuff. We've also all seen the failures, and they tend to be funny right up until they aren't.
Where it starts to concern me is narrower than the whole technology. It's the picture of somebody at midnight — smart, capable, making good decisions in every other part of their life — typing a very big financial question into a chat window because the house is quiet and it's finally been on their mind long enough to ask.
They get an answer. It's clean, it's confident, and it's simplistic in a way that's hard to see at midnight. It settles the thing enough that they can go to bed. And it may keep settling it for months, while they head down a road that isn't the right one for them.
The lost months are the part that stays with me.
Speed is most of the problem
Every one of us has more to do than time to do it in. That was true before any of this arrived. What's changed is that we now have something that produces a finished-looking answer in about nine seconds, formatted cleanly, written confidently, with no visible seams. When you have eleven other things waiting, that's very hard to argue with.
And here's what I've noticed about who this catches. It catches the capable — the people carrying the most, moving the fastest, with the least room in the day to double back. Fluency reads as accuracy. A clean answer feels like a checked answer, and the feeling is enough when you're behind.
Mistakes have always been part of the deal. What's different now is the pace. We used to be wrong slowly, with time to notice, with somebody in the chain to catch it. Now we can be wrong at the speed of the internet, in volume, and look polished doing it.
Why this matters more with money than most places
In a lot of areas, a wrong answer surfaces quickly. You follow the recipe and dinner is bad. You follow the directions and end up on the wrong street.
Financial decisions don't work that way. A withdrawal strategy that's wrong for your situation doesn't announce itself. It just quietly costs you, and you find out in year three, or year seven, or when your tax preparer calls with news. The feedback loop is long enough that you can be off course a good while before anything tells you.
Which brings me to the questions people are actually typing into these tools. I hear about them constantly now, and clients bring me printouts, which I like — I'd rather someone walk in with questions than walk in with none.
"Can I retire at 62?"
The tool will give you an answer, and it will be built almost entirely on what you've got saved. Health coverage between 62 and Medicare is the piece most people underestimate, and for some it's the single largest line item of those three years. But it's rarely the only thing sitting in the way. There's the parent who may need help sooner than anyone's admitting, and what that costs in money and in working hours. There's the adult child whose situation hasn't fully resolved. There's a spouse on a different timeline, still working, still carrying the insurance, with their own idea of when this starts. There's the mortgage that doesn't end when the paycheck does, the roof that's got about four years left in it, and the pension or Social Security decision that changes shape depending on which one of you claims first. Answer that question with a savings balance alone and you'll get a yes or a no that has almost nothing to do with your actual life.
"Should I do a Roth conversion this year?"
This one depends on a multi-year picture — what your income does between now and required distributions, what's happening with company stock, whether there's a low-income window coming that nobody's noticed yet. A single-year answer to a multi-year question will sound completely reasonable and still be the wrong move.
"Am I diversified enough?"
This is the one I'd push back on hardest, because the tool can only answer it about what you fed it. And almost nobody feeds it everything. The old 401(k) from two jobs ago is sitting somewhere. The Roth your spouse opened is on a different login. The rental property, the business interest, the HSA you've been quietly investing, the account your father left you that you haven't touched — none of that made it into the chat window. Neither did the paycheck, which for most households is the largest single asset they own and the one most tied to a specific industry. So the answer comes back about a slice, phrased like it's about the whole. Someone can look beautifully diversified in that slice and be heavily concentrated in real life. Getting this right takes a full accounting, updated as things change, and it takes somebody asking what you forgot to mention.
"How much do I need?" This is the one people search. It's also not really the question.
The question underneath
What people want to know is how much they can spend.
Because it doesn't matter what you've accumulated if you're afraid to touch it. I've sat across from people who did everything right for thirty years and then rationed themselves through a retirement they could have enjoyed. The money was there. Nobody ever connected it to permission.
"How much do I need" is a target you save toward. "How much can I spend" is a number you have to live with, and it's the one that determines whether the retirement is any good. And it isn't one number. It's the travel years and the quieter years after them. It's the help you want to give your kids while you're here to see it. It's the bridge before Medicare and the cushion for the stretch nobody plans for.
A tool can apply a withdrawal rate to a lump sum. It can't tell you which accounts to pull from in which order, what happens to the plan if the first three years go badly, or what you actually wanted this money to do. It was never told. It doesn't ask. It fills the gaps with assumptions and doesn't mention which ones.
One input among several
Portfolio analysis is a real input. So is the tax return. So is the vesting schedule, the insurance review, the estate documents, and a conversation about what you want the next ten years to look like. My job is holding all of it in the same frame and finding the places where the pieces don't line up — because that's usually where the money is being left.
The other thing I bring is time in the chair. Thousands of hours across from people in situations close enough to yours that I know what the number on the page usually means about the person holding it, and what they haven't said out loud yet.
So bring me what the tool told you. Print it, screenshot it, forward it. We'll go through it together and sort out which parts hold up once your actual life is in the room.
Use every tool you've got. Then build it on purpose.