How AI predicts the outcome of a job change

Most people decide on a job change the same way: they make a pros-and-cons list, sleep on it, and then go with the option their gut was leaning toward the whole time. The list was theater. What you actually wanted was a prediction — a clear-eyed view of how each path is likely to play out — and a list can't give you that.

That's the gap an AI prediction fills. You describe the move you're weighing, and instead of a tidy verdict, you get the likely outcomes laid out with probabilities, the assumptions behind each one, and the single factor the whole thing turns on. This post walks through how that prediction gets made, so you can run your own and read the result without fooling yourself.

Why a job change is hard to predict by feel

A job change packs several uncertain variables into one decision: the new role's actual day-to-day (rarely what the interview implied), your ramp speed, the team you can't see until you're inside it, and the opportunity cost of the role you're leaving. Your gut compresses all of that into a single feeling, and feelings are notoriously bad at weighting low-probability, high-impact outcomes — the manager who leaves three months after you join, say.

A prediction refuses to compress. It keeps the variables separate, runs them forward, and shows you the spread. The point isn't to remove the uncertainty; it's to stop your optimism from quietly deleting the bad branch.

What the AI assumes before it predicts

The first thing MiroFish does with a job-change scenario is write down its assumptions. For a move from a stable senior role to a smaller, faster company, that ledger might include: a 3–4 month ramp before you're fully productive, a 20–30% chance the role's scope drifts from what was described, equity that's only worth modeling at a steep discount, and a baseline that your current job remains available-ish (it usually isn't, once you've mentally left).

This is the part to read most carefully. If an assumption is wrong — say you can actually ramp in six weeks because the stack is familiar — you change it, and the prediction updates. The assumptions are where you inject what you know that the model doesn't.

How the outcomes branch

With assumptions in hand, the prediction splits your scenario into a handful of weighted paths. A typical job-change prediction looks something like:

• The good fit (most likely, ~45%): You ramp on schedule, the scope holds, and within a year you've grown into responsibilities the old role couldn't offer. Pay catches up; equity is a lottery ticket you don't count on.
• The lateral (~30%): Comparable pay and growth, just different. Not a mistake, not a leap — the move was emotionally significant and materially neutral.
• The misfire (~25%): Scope drifts, a key person leaves, or the culture isn't what the interviews suggested. You're job-hunting again inside 18 months, now with a short stint on your résumé.

Seeing those side by side does something a pros-and-cons list never does: it makes the misfire concrete and gives it a real weight. A 25% chance of an unhappy exit is not a rounding error you get to ignore.

The factor it usually turns on

Here's where the prediction earns its keep. The deciding factor for most job changes isn't the salary delta — it's almost always whether the role's scope survives contact with reality. People assume the money is the swing variable because it's the easy number to compare. The prediction will usually tell you the swing variable is scope clarity: roles that were described precisely and confirmed in writing land in the good branch far more often than roles sold on vibes and "you'll figure it out."

That reframes the question you bring back to the negotiation. Instead of squeezing another few thousand out of the offer, you spend your leverage getting the first-90-days scope in writing — because the prediction says that's the variable that actually moves your outcome.

Reading the prediction honestly

Two cautions. First, you have a preferred answer going in, and it's tempting to read the branches as confirmation. Discipline yourself to look hardest at the branch you don't like. Second, a job-change prediction is only as good as your inputs — if you described the new role in one vague sentence, the model invented most of it. Spend five minutes giving it the real details.

If you want the mechanics behind all this, read how AI scenario prediction actually works and how to write a scenario question to get a useful prediction. And if your job change is really a career change, the sibling post on whether a career switch pays off goes deeper on the longer horizon.

A job change is one of the few decisions where you genuinely can't run the experiment twice. That's exactly the situation a prediction is for.