A $150k employee no longer costs $150k
A $150k employee no longer costs $150k.
AI is making employee cost dynamic. Salary is only the starting point.
Two people with the same title can now have very different economics depending on the AI they consume and the useful output they produce.
That gives you four quadrants:
High productivity, low AI cost: exceptional leverage.
High productivity, high AI cost: expensive, often worth funding.
Low productivity, low AI cost: increasingly exposed.
Low productivity, high AI cost: paying twice for mediocre output.
The naive business owner sees the AI bill and reacts predictably.
“Why is this employee spending $1,800 a month on models?”
So they cap usage.
Now the employee who was producing $30k of additional value to save $1,000 becomes less productive. Procurement celebrates. The P&L quietly gets worse.
Or management goes the other direction: “Everyone needs to use AI.”
Now people generate more documents, more analysis, more code, more campaigns. Activity rises. Useful output may not.
Both reactions optimize the visible number.
The second-order consequence is more interesting: compensation, hiring, management, and budgeting all start changing when output per employee becomes less correlated with salary.
The better approach is to manage the quadrants.
Measure AI cost against useful output. Give high-leverage employees more compute when the economics support it. Investigate expensive low-output behavior. Stop treating AI usage itself as progress.
The uncomfortable truth is that a cheap employee can now be expensive, and an expensive employee can become surprisingly cheap.
Headcount × salary is becoming a worse model of workforce cost.