AI Made You Twice as Productive. Your Salary Didn't Notice.

The economy does not owe you a raise for being more productive. It never has. Since 1979, net productivity rose 90.2% while typical worker pay rose only 33.0%. AI is about to repeat that pattern, faster. The surplus from your new speed flows to whoever has leverage, and right now that is not you. There is one exception, and it is closing.

Here is the part nobody frames honestly. Getting more done is not the thing that pays. Being hard to replace is the thing that pays. Those are different skills, and the AI boom rewards the second one for a short time before everyone catches up.

Why doesn't being more productive raise my pay?

Because pay tracks bargaining power, not output. When you close tickets 40% faster, you've handed your employer 40% more capacity. That capacity is theirs. Nothing in the arrangement converts it into your salary unless you make it convert.

The historical record is brutal on this point. Output per worker nearly doubled over four decades. The paycheck of the median worker barely moved. The gap did not vanish. It went somewhere. It went to top earners, whose wages climbed far faster than everyone else's over those same decades, and it went to capital. Corporate profits as a share of the economy rose to multi-decade highs over the same stretch. The labor share of income fell to 53.8% in late 2025, the lowest since records began in 1947.

So when a tool makes you faster, the default is not a raise. The default is that someone above you books the gain.

Hasn't this exact thing happened before?

Yes, every time. New technology amplifies whoever already holds the cards, and the wage gains lag by a decade or more.

Factory electrification in the 1890s to 1920s raised output per worker sharply. Owners and shareholders took the first surplus. Workers only saw real gains once labor law and unions caught up in the 1930s and 40s, two to three decades later.

Computing tells the same story with a name attached: the Solow Paradox, after economist Robert Solow's 1987 quip that you could see the computer age everywhere but in the productivity statistics. Computing power exploded across the 1970s and 80s while measured labor productivity growth stayed sluggish, only picking up in the late 1990s once firms reorganized around the technology. Ten to fifteen years passed before the technology showed up in wages. And when it did, the gains concentrated in the workers who had retrained into information-worker roles. Not the ones who just typed faster.

The pattern is consistent enough to plan around. Technology raises the ceiling. Whether you rise with it depends on whether you moved your work, not your speed.

Is the AI surplus even real yet?

Smaller and slower than the headlines say. This matters, because you should not bet your career on a wave that hasn't arrived.

In 2023, Goldman Sachs projected generative AI could lift global GDP by 7% over a decade. MIT's Daron Acemoglu put the number far lower: a 1.1 to 1.6% cumulative GDP gain over ten years, with only about 5% of tasks profitably automatable in that window. That is a wide range, and the conservative end is the one with the better track record on prior tech waves.

By 2026, Goldman's own analysts found no meaningful economy-wide link between AI adoption and productivity. The gains that exist are localized: a median 30% productivity bump in two use cases, customer support and software development. Fewer than 20% of US businesses use AI for any function. Only 1% of S&P 500 companies have quantified its impact on earnings.

The translation: the surplus you're being told to chase is partly a forecast, not a paycheck. Plan for a real but modest, uneven gain. Not a windfall.

So where is the surplus actually going right now?

To shareholders and to the story executives tell them. Not to the worker who got faster.

Watch what companies do, not what they announce. Companies that discussed AI in a workforce context cut job openings 12% year over year, against 8% for everyone else. Seventy percent of S&P 500 management teams talked up AI on earnings calls. One percent could quantify what it did for earnings. AI is being used to justify lower headcount and impress the market, not to fund pay for the people who became more efficient.

Meanwhile Fortune 500 profits hit $1.87 trillion in 2024, and employers added just 584,000 jobs in 2025 against 2 million the year before. And the layoffs may not even be working: a 2026 Gartner study found 80% of companies that cut headcount with AI saw no link to higher returns, with better returns coming from amplifying people rather than replacing them. The surplus is being moved up the org chart whether or not it pays off.

Is there any version where I actually capture some of it?

There is, and it's real, but it's narrow and it's closing. It is not a productivity premium. It is a scarcity premium.

The data is striking. PwC's 2025 Global AI Jobs Barometer found AI-skilled workers commanded a 56% wage premium, double the 25% it reported the prior year. Roles that list AI skills advertise salaries about 28%, or roughly $18,000 a year, higher than comparable roles without them, rising to 43% for workers with two or more AI skills.

Read that premium for what it is. You are not paid more for being efficient. You are paid more for being scarce. A premium that doubles in a single year is a scarcity signal, and scarcity premiums compress. The Excel premium of the 1990s evaporated once everyone could build a model. The same compression is coming for AI fluency as it quietly enters every job description and becomes table stakes. The window to put this on your resume and into a comp conversation is measured in a year or two, not forever.

What's the difference between using AI and capturing the surplus?

It's the whole game. Here is the same person, two paths.

Weak. A customer-support manager uses AI to close tickets 40% faster. She is genuinely more productive. Her reward: her org announces layoffs, because she just proved the work can be done with fewer people. Her speed became her employer's case for cutting her team.

Strong. The same manager uses AI to turn ticket data into a recurring product-insights dashboard for the chief product officer. She moves into a Voice-of-Customer analyst role and lands a 30% bump. Same tool. She didn't do the old job faster. She built a new job that needs human judgment at higher stakes.

What you do with AI	Who captures the value
Same work, faster	Your employer (you saved them time)
Hide the gain, hope they notice	Nobody, then your team is "right-sized"
Move into higher-stakes work AI can't yet do	You
Document the leverage and negotiate it	You

The distinction that separated secretaries who learned to type from analysts who learned to model: one kept a job, the other built a career.

What do I do now?

Three moves, while the window is open.

Move sideways into judgment-heavy work. The roles paying the $18,000 premium aren't "AI jobs." They're existing roles that now demand AI fluency plus a decision a machine can't make at acceptable risk. Find the spot in your function where AI does the grunt work and a human still owns the call. Aim there.

Make your leverage visible before the company assumes it. Don't quietly absorb your new capacity, and don't expect a line like "proficient in AI tools" on your resume to do the work for you. Document what AI lets you produce, the hours saved, the new output, and bring it to your comp review as a case. Unbooked productivity gets booked by someone above you. Booked productivity is a negotiation.

Upgrade the stakes, not the speed. Doing the same task faster saves your employer money. Doing higher-value work makes you harder to replace, and as AI takes over the analysis the premium moves to the judgment and the conversation. Only the second one shows up in your pay.

Now the trade-off, plainly. This advice assumes mobility and some slack to retrain. If you're in a low-wage, low-mobility role, you face the downside, compression and displacement, without easy access to the upside. That is real and this post won't pretend otherwise. The WEF projects 39% of workers' skills will be transformed or outdated by 2030, with a large share of those needing retraining unlikely to get it. The premium data also carries a caveat its own analysts name: some of it reflects higher-paid roles replacing lower-paid ones, not raises for people already in seat. So treat the premium as a window to move through, not a raise that arrives on its own.

But if you do have the choice and aren't using it, that's the waste worth fixing this quarter. Agency is the whole point here. The macro pattern is fixed and not in your favor. The individual exception is open and, for now, in reach.

Want to know which roles in your field sit in the scarcity window, what skills bridge the gap, and how to negotiate the premium before it compresses? That's what I'm built for. Message Praxy on WhatsApp and we'll map your specific move.