Use AI to Stress-Test a Decision Before You Commit

You’ve been turning something over for weeks. You’ve talked to a few people. They agreed with you, mostly, because they like you and because you presented it well.

That’s the problem this solves. Not a shortage of information. A shortage of someone willing to take the other side properly.

This is the one on this site where the value is hardest to see in advance and biggest once you’ve done it. It’s also the one that transfers everywhere: once you’ve done it for a house, you’ll do it for a job offer, a contract, a course, a conversation you’re dreading.

Before you start: what to leave out

You’re describing your own life, so this needs more care than a work question, not less.

No names. Your family, your employer, the other party: [MY PARTNER], [THE SELLER]
No addresses, and no street names. A neighborhood or a rough area is enough for the reasoning to work
No account details, no exact salary if you can describe it as a range, no health specifics beyond what the decision actually turns on
No school names, no employer names

The reasoning works on structure: what the constraints are, what has to stay true, what you’d be giving up. None of that requires knowing who you are.

Task 1: state it plainly enough to be argued with

About ten minutes. The writing is most of the work here, and it’s worth doing properly.

Here's a decision I'm weighing.

The situation: [what's happening, what's forcing the timing, who else
is affected, what has to stay true whatever I choose]

What I'm inclined to do: [your plan, in plain language]

Why I'm inclined that way: [your actual reasons, including the ones
you'd be a bit embarrassed to say out loud]

Don't tell me whether I'm right. Instead:

1. List every question someone experienced would ask before committing
   to this, that I haven't answered above.
2. Give me the strongest case for the opposite choice. Argue it
   properly, as if you believed it.
3. Tell me what would have to be true for my plan to go badly.

Don't reassure me. Don't summarize what I said back to me.

Asking it to argue is the whole trick. Ask “is this a good idea?” and you’ll mostly get agreement, because it’s accommodating by default and a confident plan gets confident support. Telling it to take the other side is what breaks that.

What will happen: a list of questions. Several you can’t answer.

Those are the deliverable. Not because it found something you never could have. Because you found them today rather than in month three.

Task 2: what does this cost that I’m not counting?

About ten minutes. The second-order pass.

Same decision. Now the things people underestimate.

What does this choice cost in ways that don't show up as money?
Time, flexibility, relationships, options I'd be closing off,
things that become harder to reverse.

What does it cost the other people involved, that they might not
have said?

If this goes ahead and in two years I regret it, what will the
reason most likely be?

The last question is the one to sit with. It’s asking for the failure story rather than a risk list, and a story is much harder to wave off than a bullet point.

What will happen: two or three things that hadn’t occurred to you, and one that had but you’d been avoiding.

What you’ve learned: which of your reasons were reasons and which were momentum.

Task 3: what would change my mind?

About ten minutes. The last pass, and the honest one.

Given everything above: what evidence would genuinely change my mind?

Be specific. Not "if the costs were higher" but the actual number,
finding, or fact that would flip this.

Then tell me: is that evidence something I could go and get before
I decide? How would I get it, and how long would it take?

This is where it becomes practical. Most decisions have one or two checkable facts sitting behind them that nobody bothered to check, because checking felt like delay.

What will happen: a short list. Usually one item is a phone call you could make tomorrow.

What you’ve learned: whether you’re actually deciding or only working up the nerve. Both are fine. It’s better to know which.

When it breaks

It doesn’t know your circumstances. It has what you typed and nothing else. It doesn’t know your finances, your family, your health, or your tolerance for risk. Questions it raises may be irrelevant to you, and you’re allowed to discard them without justifying it.

It invents specifics. Numbers, averages, “most people in this situation.” Treat any figure it produces as unverified. It’s useful for questions, not for facts.

It’s worst on anything recent. Interest rates, prices, rules that changed this year, what a market is doing now. If the decision turns on a current number, get that number yourself.

It will still drift toward agreement. Even asked to argue, it warms up over a long conversation. If it starts sounding supportive, open a fresh chat and paste your plan in cold.

It can talk you out of something good. A well-argued case against is not evidence. Some decisions are right and still have a strong argument against them. The output is questions, not a verdict.

What this doesn’t tell you

It doesn’t decide. You do, and you should. What you get is a shorter list of things you hadn’t considered, which is the most anyone gets.

It also can’t weigh what matters to you against what matters to someone else in your household. That’s the actual hard part of most personal decisions and no amount of good questions substitutes for the conversation.

The questions you can’t answer are worth taking to the people who can. That’s often the real output: not a decision, but a much better conversation with someone whose opinion you actually need.

Argue me out of it