I have a long history of being less-than-great with money.

For decades, I developed an almost impressive ability to talk myself into spending it—and around nearly every reasonable objection that might have gotten in the way.

I used to think this meant I simply needed to grow up. Be more disciplined. Exercise more willpower.

I don’t really believe that anymore.

As I’ve gotten older, I haven’t necessarily stopped wanting to spend money, though I have definitely become more aware of the systems that produce that savings-phobic impulse. Not the external systems that are easy to blame—marketing, advertising, and their evolution into something shockingly personalized, targeted, and engaging.

The real systems, the internal ones.

The more I’ve read about how we actually make decisions, the more I can see the shape of that system—and the more obvious it becomes that I’ll never have a clean read on all of its inputs at any given moment.

Biology, developmental history, reinforcement, environment, present emotional state, available information, defaults, friction. Countless internal and historical conditions are influencing the state in which any particular decision gets made.

Mine, I’ve learned, likes to spend money. On things, on experiences, on ideas.

None of those are bad, of course. They are some of the things—not all of the things—that enrich our lives. But when the governor isn’t doing much governing, the wheels can come off. Which can often restart the loop, digging further downward.

For nearly as long as I’ve recognized this as something I needed to get better at, which is to say most of my adult life, I’ve felt personally responsible. Like I had failed to be a grown-up, or a man, or a boyfriend, or a father.

And hey, those things aren’t not true. At least on occasion.

But once I started to become more aware of the complex conditions that built that system and continue to feed it, I realized there was more going on. Some of those conditions simply make certain decisions easier for me than others.

That doesn’t remove my ownership of the consequences. It gives me more places to intervene.

For years, I’d wished I could just hand my bank account over to someone and tell them to manage everything for me. Give me an allowance. Give me guidance on the big splurges and tell me if they’re good moves, maybe reasonable in another paycheck or two, or if I just need to let those last three mai tais wear off before we talk about this again.

But hey, even if those professionals exist, I can’t afford them. And even if I could, they wouldn’t be there at 11 p.m. on a Saturday night when I’m at the bar thinking I should buy those show tickets and book that nice hotel in Chicago for next weekend.

Then I remembered the always-on assistant in my pocket: ChatGPT.

I’d been skeptical of connecting my financial accounts. That felt like a lot of information to hand over to OpenAI. It honestly still does. But my discomfort with that has been eclipsed by my discomfort with not growing my bank account alongside my income.

And now I have something much closer to what I was actually looking for.

Of course, this isn’t a perfect system.

ChatGPT can be wrong. I can misunderstand myself, broadly or in the moment. It can only know what I’ve told it and what the data it has available can show. I still have to be the one to actually open the app, ask the question, and then consider what it tells me.

But the system it’s supplementing is also imperfect.

How many times have you made a decision and almost immediately said to yourself, “I knew better than that”?

That information existed. You knew your goals. You were otherwise capable of making a smart, informed decision. But that’s not what your system optimized you for in that moment.

I like to think of ChatGPT as another sanity check and pressure valve in my system. One I’ve spent dozens of hours talking to, that knows a surprising amount about not only what I am trying to accomplish but why. I still have to invoke it, and I can still ignore it.

But it’s there, in a place I am already accustomed to looking, when I need access to a version of my own thinking that isn’t quite so impressed by whatever thing is calling my name and demanding my money.

ChatGPT has helped me build a path forward that looks, to me, realistic and sustainable using my actual data. My actual habits. My actual goals.

I’m not asking ChatGPT to decide what I value. It didn’t choose my goals. It doesn’t decide what kind of life I want.

What it can do is change the conditions under which I make decisions. It can put friction between impulse and action. It can remind me of information that is very easy to forget when I’m excited about something. It can help me see whether the story I’m telling myself about a purchase still makes sense when placed beside the goals I’ve already said matter more.

In other words, it can make relevant information harder for me to conveniently forget.

Cognitive scientists have a term for a version of this: cognitive offloading. We’ve always used things outside our heads—notes, calendars, reminders—to reduce how much information we have to reliably carry around inside them. This feels like a more conversational version of the same move.

I use ChatGPT similarly when I’m stuck at the beginning of an idea but can’t quite yet figure out what I’m trying to say. I’ll take that half-formed idea into a kind of whiteboarding session. Not so it will tell me what to think or come up with the idea for me, but to ask questions, challenge assumptions, and help me turn that rough idea into something more fully formed.

It’s not that the AI is thinking for me. It’s helping me surface my own thinking into something I can actually work with.

Researchers are starting to explore that distinction directly. A recent study with the almost suspiciously appropriate title “AI, Help Me Think—but for Myself” compared AI that simply recommended decisions with AI designed to build on a person’s own reasoning. The latter integrated better into participants’ thinking and produced slightly better outcomes.

That’s a small study, not proof of some universal rule. But I think the distinction matters.

There is real risk that AI could empower more and more people to stop thinking for themselves. I share that concern.

But I think we’re making a mistake when we treat every use of AI as movement in that direction.

Sometimes the problem isn’t that a person refuses to think. Sometimes they already know where their judgment tends to fail. They know the goals they want to pursue. They know the kind of life they’re trying to build.

What they lack is a reliable way to bring all of that information into the moment when it matters.

That’s an important distinction.

The point isn’t that AI somehow possesses better judgment than I do. It’s that it can make the judgment I have when I’m thinking clearly harder to lose access to when the conditions change.

That doesn’t feel like replacing thought.

It feels like changing the conditions under which better judgment is more likely to show up.