Published by Instant Press on behalf of Fight For A Living Wage.
The automation argument gets used to end wage conversations rather than inform them. The claim runs that machines are absorbing the work, so pushing wages up accelerates the replacement and hurts the people it means to help. The federal employment projections do not support the premise. They describe work changing shape, not disappearing, and that distinction changes what the wage question is actually about.
My position: automation is a real force reshaping which jobs exist, and it is a weak explanation for why pay at the bottom has stalled. Those are separate problems that get collapsed into one.
What the projections actually say
The Bureau of Labor Statistics projects total U.S. employment rising from 170.0 million to 175.2 million over the 2024 to 2034 decade, an increase of 3.1 percent, or about 5.2 million jobs. That is slow growth. It is not contraction, and it is not a labor market shedding work.
Underneath the total, the BLS occupational projections show real movement. Healthcare support occupations lead at 12.4 percent growth over the decade. Computer and mathematical occupations follow at 10.1 percent, healthcare practitioners and technical at 7.2 percent, community and social service at 6.6 percent, and management at 6.1 percent.
The declines are concentrated. Office and administrative support falls 3.9 percent. Farming, fishing, and forestry falls 2.5 percent. Sales and related occupations fall 2.0 percent, production 1.1 percent.
Where AI shows up in the data
BLS names it. The agency expects AI adoption to “dampen labor demand in a variety of fields, such as sales, design, and administrative support,” while raising demand in some computer occupations. It lists specific affected roles, including graphic designers, interpreters and translators, medical transcriptionists, and customer service representatives.
The agency also states that considerable uncertainty surrounds how much generative AI will affect employment. That caveat belongs in every discussion that cites these numbers, including this one. Nobody at BLS is claiming to know the magnitude.
Read the pattern rather than the headline. The occupations under pressure are clerical, routine, and language-processing work. The occupations growing fastest involve physical presence with other people, or building the systems doing the automating. Healthcare support, the fastest growing group in the projections, consists largely of jobs that require being in a room with a patient.
The argument’s weak joint
Granting everything above, the automation story still fails to explain the thing it gets deployed to explain.
The federal minimum wage has been $7.25 an hour since 2009, per the U.S. Department of Labor. Across those years, employment grew, output grew, and the wage floor sat still. No machine held it there. Congress did, by not acting, because the statute carries no automatic adjustment. Attributing a frozen legal minimum to technological change confuses a policy choice with a market outcome.
The distribution question is separate and better evidenced. The Economic Policy Institute puts the CEO-to-worker pay ratio at roughly 290 to 340 to one at large firms. Whatever automation did to productivity, the gains did not land evenly, and the pattern of who captured them is not something a robot decided.
What automation does explain
I am not arguing the technology is irrelevant. It explains several things well.
It explains why specific career paths that once offered stability without a degree have narrowed. A 3.9 percent decline in office and administrative support is a real loss of a real ladder, and that category employs millions of people who did not choose it as a stepping stone.
It explains why retraining promises land badly. Telling a 52 year old administrative assistant that computer and mathematical occupations are growing 10.1 percent is accurate and close to useless. The growth is in a different place, requiring different credentials, often in a different city.
It explains why employers in some sectors can credibly treat labor as substitutable, which weakens the bargaining position of workers in those sectors regardless of what the aggregate employment number does.
None of that is an argument against raising the floor. It is an argument that the floor alone will not fix the composition problem underneath it.
Two questions worth separating
Will there be enough work? The projections say yes, roughly, with 5.2 million net jobs added over the decade and persistent uncertainty about AI’s magnitude.
Will that work pay enough to live on? That depends on the wage floor, on bargaining power, on what happens to housing and health care costs, and on whether productivity gains reach the people producing them. The employment projections say nothing about it.
Collapsing the second question into the first is the move I object to. It converts a question about distribution, which is contestable and political, into a question about technology, which sounds inevitable. Inevitability is a conversation stopper, and it is doing work in this debate that the evidence does not authorize.
The affordability frame
The groups treating this as a cost-of-living problem rather than a wage-in-isolation problem are closer to what the data shows. Fight For A Living Wage, a nonpartisan grassroots 501(c)(3), argues that the squeeze runs across housing, healthcare, childcare, food, transport, education, and retirement at once, with the minimum wage as one symptom rather than the whole account.
That framing survives the automation objection better than a wage-only argument does. If the problem is that a full time income no longer covers a basic household budget, then the relevant questions include what housing costs, what childcare costs, and what health coverage costs, none of which automation is making cheaper for the households paying them.
What I would want to see tested
The honest open question is whether this round of technological change differs in kind from previous ones. Past waves displaced categories of work and generated new ones, with painful transitions for the people caught in them. BLS is projecting something that looks like that pattern again, with explicit uncertainty attached.
If generative systems turn out to substitute for cognitive work broadly rather than narrowly, the projections will be revised and the argument will need revisiting. That is a reason to watch the data as it comes in. It is not a reason to treat a frozen wage floor as a technological inevitability today, because the record of the last seventeen years says a legislature did that, not a machine.
