So My Job Got Automated and Everyone Kept Saying "Now You're Free"

I got the email on a Tuesday. The trend tool had been right about three seasons in a row and the math stopped working for a human whose main value was knowing what would sell in eighteen months.

Here's the thing nobody tells you about being replaced by a prediction model: it's not dramatic. There's no scene. You just gradually stop being in the room where the decisions happen, and then one day you're not on the calendar, and then you're having a very polite conversation with someone from HR who keeps saying "restructure."

I spent about six weeks being unbearable about it. Then I got bored of my own tragedy and started looking at what the tool actually bought and what it actually missed, because I have a very specific set of skills and one of them is knowing the difference between a trend and a moment.

Four seasons later, I'm still in fashion. Different job. Better job, actually, which I hate admitting because I wanted the story to be about how machines can't do taste.

What the prediction model gets right

Let's be fair to it, because being unfair to it is how you end up irrelevant.

It reads the runway shows, the resale listings, the search volume, the influencer posts, the factory orders, and it finds the pattern before a human room does. It's fast and it's cheap and it doesn't get emotionally attached to a silhouette because it had a good season with that silhouette in 2019.

Specifically, it's excellent at three things:

Volume prediction. How much of a thing to make. This used to be the single most expensive judgment in the business. Get it wrong and you're either sitting on inventory or leaving money on the floor. The tool is better than the buyers at this, and the buyers know it.

Cross-market lag. A trend that's six weeks ahead in Seoul and four months behind in Lagos. Models are extraordinary at tracking those offsets. Humans track two or three markets. The model tracks forty.

The obvious call. The thing that's going to be everywhere and everyone can see it coming. The model nails this, and honestly, so did we. It's not a high-value skill, it was just a job.

What it can't do, and I need you to understand the mechanism

The model is trained on what sold. Which means it can only predict what will sell next, extrapolating from what sold before.

What it cannot do is invent the thing that makes last season look stupid. Because that thing, by definition, was not in the training data. Nobody was buying it. That's what made it new.

The pattern I've now watched three times: the tool predicts a safe extension of the current mood. Someone with taste and no respect for the tool makes the opposite thing. The opposite thing looks insane for one season and then everything is that. Then the tool catches up and predicts extensions of the new thing, and the cycle repeats.

Taste is not predicting the trend. Taste is being willing to be wrong in public for about nine months.

No model is willing to be wrong for nine months, because that's not a thing you can optimize for. It's a thing you can only do if you're a person with a point of view and a stomach.

What I actually do now

I run a small studio. Three people. We don't do trend forecasting at all. We do the opposite: we take a small number of clients and we make one specific thing they'd never have made on their own, and we make it because one of us believes in it.

The prediction tools are still in the workflow. We use them for volume, for lead times, for the boring parts of production planning. They've taken over the part of the job that was always secretly about math.

What they haven't taken over is the room where three people argue about whether a collar is too much. That room is the whole job. It always was. We just used to fill it with forecasting meetings that felt like work.

The thing I want to say to people who are next

You're going to be told that your displacement is an opportunity, and it's going to make you want to break something, because it's said by people who aren't displaced.

But there is a version of this that's true, and it's less inspiring and more useful.

The tool will take the part of your job that's measurable. That's the part where the value was aggregate, average, predictable. What's left is the part that's idiosyncratic: the relationships, the taste, the willingness to bet, the willingness to be wrong, the ability to walk into a room and read it.

That part was always the actual job. The measurable part was how you got paid for it.

The problem is that the measurable part was also 80% of your calendar, and now that it's gone you have to actually do the other thing, in public, with your name on it, and there's nowhere to hide.

Most people won't. Not because they can't, but because the other thing requires you to have a point of view, and having a point of view is uncomfortable and public and occasionally humiliating.

The tool gave you the time. It did not give you the nerve.

Wear the future or get worn out.