I Got the “AI Skills” Raise. It Was 4%. The Tool That Replaced My Old Job Costs $240 a Seat.

They told me to upskill. Then I ran the math on who actually captured the value — and it was not me.

Photo by Sasun Bughdaryan on Unsplash

The raise came through in March. Four percent. My manager delivered it like a gift, which, in a year when half my industry was getting walked out, it sort of was. “This reflects how much more valuable you’ve become,” he said, “now that you’re AI-fluent.”

I did the thing you are not supposed to do with a compliment. I took it home and ran the numbers on it.

Four percent of my salary is a specific amount of money. I am not going to print my salary here, but I will tell you the raise worked out to a little less than I spend on coffee in a year. It was real. I was grateful for it. And it was the smallest part of a transaction I was only beginning to understand.

Because here is the other number. The AI tool I had become so admirably “fluent” in — the one that now does the bulk of what my old job used to be — costs the company somewhere around two hundred and forty dollars a seat, per month, on the enterprise plan. There are a lot of seats. I know roughly how many, and I know roughly what we bill clients for the work that tool now accelerates, and the gap between those two figures is not four percent. It is not anywhere close to four percent.

I had been handed a raise framed as a reward for becoming more productive, in a year when “more productive” was doing a tremendous amount of quiet work.

The studies, when I looked, were almost comically consistent. Workers using AI tools were reporting productivity gains of twenty to sixty percent on the tasks the tools touched. Twenty to sixty percent. And the wage data? A few points here, a cost-of-living bump there, a four percent “AI-fluency” raise delivered like a bonus. The value was real and it was large. It simply was not landing in my account. It was landing somewhere upstream, in a line item I would never see, captured by people who had understood something I was slow to learn: that producing value and capturing value are two different jobs, and only one of them pays.

There is a phrase economists started using around this time — a “two-track” labor market. On one track, AI makes certain workers more valuable: the ones whose judgment, relationships, and tacit knowledge the tool amplifies rather than replaces. Their wages rose, in some studies meaningfully. On the other track, AI does the opposite. It takes a job that used to require a skilled person and turns it into a job that requires a person who can operate the tool — which is to say, a more interchangeable person, a cheaper person over time. Same tool. Opposite effect. The whole question is which track it puts you on.

And here is the trap I had walked into without noticing: the “AI skills” raise is what the second track feels like from the inside. It feels like winning. You learned the new thing. You got rewarded. You are, technically, more productive than you were last year. But “AI-fluent” can mean two very different things. It can mean you became more valuable. Or it can mean you became the person who feeds the machine that made your old expertise optional — and got handed four percent to feel good about it.

I am not sure yet which one I am. That is the honest and uncomfortable center of this. Some days I think my judgment is the thing the tool cannot replace, and I am on the good track, and the raise was just slow to catch up. Other days I notice how much of my week is now spent prompting, checking, and cleaning up after a system that does the part I used to be paid for, and I wonder whether I have been promoted into my own obsolescence with a four percent ribbon on it.

What I am sure of is that the framing was doing work. “You’ve become more valuable” is a sentence engineered to make you stop running the numbers. It flatters you onto the captured side of the ledger and calls it a raise. The genuinely valuable move — the one the company made and I almost missed — is to make the person generating the value believe they are the one being rewarded, while the value itself flows quietly somewhere else.

So I have started asking a different question at review time. Not “did I get a raise” but “of the value I helped create this year, what share did I capture?” It is a worse question to ask. It does not feel good. It will probably make me harder to manage. But four percent of my salary against twenty-to-sixty percent of my output is not a number I can unsee, and I would rather know which track I am on than be comfortable about it.

The raise is in my account now. I am keeping it. I am also keeping the spreadsheet — the one with my four percent on one line and two hundred and forty dollars a seat on another — because the next time someone tells me I have become more valuable, I want to be able to ask the only question that matters: valuable to whom?