Comforting slogans travel fast. “AI won’t take your job” is one of them. It implies stability and continuity, and assumes the labour market will adjust slowly.
It will not.
When Jack Dorsey announced that Block would cut roughly 40 percent of its workforce, the company was not collapsing or fighting for survival. Revenue was growing. Profitability was improving. The reduction was framed as structural, enabled by intelligence tools and smaller, flatter teams. In practical terms, the output per employee had shifted enough to justify fewer employees across the organisation.
That is repricing.
Over Hiring or Structural Shift
Some argue this is simply a correction. Silicon Valley has a habit of expanding headcount during cheap capital cycles, building inflated teams, experimental roles, and redundant management layers. Twitter was frequently cited as a prime example of that expansion culture.
Then Elon Musk acquired the company and executed aggressive cuts, removing a substantial share of the workforce in a compressed timeframe. Predictions of systemic failure were immediate.
It did not happen.
Twitter, now operating as X, continued to function. The platform remained live. Core systems stayed intact. Features continued to ship. In some areas, velocity increased despite the smaller team. There was turbulence. There were trade offs. But the assumption that deep cuts would render the company unworkable proved incorrect.
It is convenient to frame both Block and X as over staffing being trimmed back to rational levels. That explanation feels temporary. It implies equilibrium returns once excess is removed.
Maybe.
Or maybe those staffing levels were viable only because cognitive work required more human input at every stage. When intelligence tools compress execution time, automate drafting and analysis, and increase the leverage of a single capable operator, the threshold for “over staffed” shifts downward.
What looked appropriate three years ago now looks expensive.
The Economics of Cognitive Compression
Salaries reflect scarcity.
For years, experienced analysts, engineers, and operators commanded high compensation because their cognitive output was difficult to replicate. Knowledge accumulated slowly. Execution required coordination across multiple roles. Human bandwidth was the constraint.
AI changes that constraint.
When one well trained individual can coordinate models to draft reports, analyse datasets, generate code, test variations, and synthesise insights in hours rather than days, the effective supply of cognitive output increases. If supply increases while demand remains relatively stable, price adjusts.
Not symbolically. Practically.
The job may still exist. The function may remain necessary. But if the same workload can be delivered with fewer people, the market value of each additional role declines. Organisations respond to that arithmetic quickly.
This is why the language of correction feels incomplete. Correction implies a temporary distortion. Repricing implies a new baseline.
What This Means for You
The uncomfortable reality is that many roles are not disappearing. They are being evaluated against a new productivity benchmark.
If your output can be materially amplified through AI tools, your leverage increases. If your role consists primarily of repeatable cognitive tasks that models now perform competently, your bargaining power weakens. The difference is no longer experience alone. It is integration.
Companies are not making moral decisions about headcount. They are making economic ones.
You can call it poor hiring. You can call it restructuring. But the pattern is consistent. When output per person rises structurally, headcount requirements fall. When headcount falls, labour pricing adjusts.
Your job may remain but its valuation is what is changing, so act accordingly.
Get comfortable using AI tools daily, not occasionally. Learn how different models behave under pressure. Understand their limits. Understand their failure modes. Know where they hallucinate and where they accelerate you. Treat them as operational infrastructure, not novelty software.
Structured learning matters. Take formal AI training where it is relevant to your function. Build internal workflows around models rather than using them ad hoc. Measure your own output before and after integration. Increase your personal leverage deliberately.
The market is setting a new standard for productivity. You can either be priced against it, or you can operate at it.
That choice will always be yours.