Wages Up, Jobs Down
What was hiding in Friday's labor report
On Friday, we learned that the U.S. economy lost 92,000 jobs in February. Economists expected a gain of 50,000. December was revised to a loss of 17,000. With those revisions, 2025 recorded five months of labor market contractions… the most since 2010.
Labor force participation dropped to 62%, its lowest since December 2021. Long-term unemployment hit 25.7 weeks, a four-year high.
And yet despite all of that, wages rose. Up 0.4% for the month and 3.8% year over year, both above expectations.
Fewer jobs. Higher pay. Those two things aren’t supposed to happen at the same time.
Let’s look at what’s actually going on.
The Quiet Split
Information services, the sector most directly exposed to AI, has lost jobs for twelve consecutive months, averaging 5,000 per month. Manufacturing lost 12,000. Federal government employment dropped 10,000, part of a 330,000-job slide since October 2024.
Meanwhile, research from the Federal Reserve Bank of Dallas found that wages in AI-exposed sectors are rising fast. Since late 2022, average weekly wages in computer systems design are up 16.7%. The national average is 7.5%.
AI isn’t just eliminating jobs. It’s starting to split the labor market in two.
Codifiable vs. Tacit
The Dallas Fed draws the line clearly. Codifiable knowledge, which is rules, procedures, templates, processes, is the kind of work that follows a pattern. AI can learn it fast and efficiently. Tacit knowledge, which is judgment, context, and the ability to navigate ambiguity, comes from experience. AI can’t replicate it.
We are seeing that workers with codifiable skills are being squeezed out. On the other hand, we see workers with tacit knowledge are commanding premiums. The same technology is automating one group and making the other more valuable.
The One-Dimensional Worker, in Real Numbers
Last week we wrote about the one-dimensional worker, the person organizations trained to execute tasks without developing judgment, adaptability, or a growth mindset. The kind of worker a machine can replace. There’s a LOT of them in enterprise-level organizations right now.
Friday’s report is that thesis in real numbers. The workers losing ground aren’t random casualties. They’re the ones whose value was built on doing predictable things predictably well. The ones commanding wage premiums bring something a model can’t: experience, relationships, and the ability to make decisions when the data is messy, and the stakes are high.
Organizations that spent years rewarding compliance and narrow execution built workforces full of codifiable knowledge. They optimized for efficiency and got brittleness.
The Compensation Problem
As we’ve mentioned before, as AI continues compressing headcount the compensation model most companies are running also breaks.
Traditional structures were designed for large workforces doing standardized work. Pay bands. Salary ranges. Annual percentage increases. Equity spread thin. That model made sense when scale meant headcount and roles were roughly interchangeable within a level.
It doesn’t work when a team of eight produces what fifty used to. The person who combines deep experience with AI fluency isn’t a 10% raise above the person who can’t. They’re a different category of contributor. And if you’re paying them like a line item in a traditional comp structure, someone else will pay them what they’re actually worth.
The wage data is already showing this. AI-exposed sectors paying more than double the national wage growth while shedding headcount. Organizations that don’t rethink how they compensate and retain their most valuable people will watch them leave. We can’t stress this enough.
The Real Report
Friday’s jobs report tells you the economy lost 92,000 jobs. It doesn’t tell you which ones are coming back.
We believe that there’s an opportunity in this data. The organizations that come out ahead won’t be the ones cutting headcount and calling it AI strategy. They’ll be the ones who recognize that the value of their people just went up, and build cultures, structures, and compensation models that reflect it.
That means investing in the tacit knowledge AI can’t touch. Developing people who can exercise judgment, not just follow processes. Creating environments where growth isn’t a perk, it’s the operating system.
For the past decade, we’ve been helping organizations build cultures where people develop, adapt, and become the kind of talent that no technology can replace. And right now, we’re building toward something new that we think will change how organizations understand their people in this moment. More on that soon.
The paradox isn’t really a paradox. It’s the market pricing in what we’ve been saying: the human advantage isn’t about doing more. It’s about knowing things a machine never will.
The companies that bet on people will win this era. The data is already proving it.
Links:
· The February jobs report: Payrolls fell by 92,000 — CNBC’s full breakdown, including the twelve-month trend in information services losses.
· AI is simultaneously aiding and replacing workers, wage data suggest — The Dallas Fed research on codifiable vs. tacit knowledge. The most important piece of economic research on AI and work we’ve read this year.
· Companies are laying off workers because of AI’s potential — not its performance — HBR on why the job losses are real even though AI hasn’t fully delivered on its promises.