Washington added developers while cutting customer service

The two occupations everyone names moved in opposite directions

Software development and customer service come up in almost every conversation about AI and work. In Washington's most recent two years of wage estimates, they went opposite ways. Software developers went from 91,470 to 107,030. Customer service representatives went from 44,880 to 39,760.

Washington now reports more than 107,000 software developers, which is a lot of people for an occupation frequently described as disappearing.

• BLS technical notes for these estimates

Computer programmers are the sharpest move

Programmers fell 21%, from 3,220 to 2,540, and their median pay fell about 21% as well. Employment and pay dropping together is unusual and worth noticing.

It is also worth knowing that programmer and software developer are separate job codes describing different work, and the programmer category has been shrinking for years as employers reclassify people into developer roles. Some of that 21% is almost certainly relabelling rather than job loss.

• How occupations are classified

Why this cannot settle the argument

The federal agency that produces these estimates says plainly that it does not encourage using them for time-series analysis. Occupation definitions get revised, the sample pools three years of responses, and older wage data is adjusted forward using assumptions.

That last detail matters more than it sounds. Because the sample pools three years, a real change in the job market shows up gradually rather than all at once. A single year-over-year move can reflect survey mechanics as easily as hiring.

So treat these as a description of what the survey recorded, not a measurement of what AI did. The direction is interesting. The precision is not there.

• Why BLS discourages year-to-year comparisons

What would be more convincing

The stronger evidence looks at age rather than occupation totals. The Stanford work that gets quoted most found its effect specifically among workers aged 22 to 25, while employment for experienced workers in the same jobs held steady.

That shape is useful because it is harder to explain another way. A soft economy tends to hit an occupation broadly. Something that mostly removes the entry rung looks different.

Washington's wage survey does not break employment out by age, so it cannot test that here. Worth keeping in mind when a state-level number gets used to argue a national point.

• Washington labor market information