The $6 Trillion Nobody Counts: What the AI Doom Debate Is Costing Working Families
The loudest story is not the important one
The loudest story about AI right now is that it ends the world. An Anthropic researcher left the company and said so. CEOs of the largest AI firms followed with letters and open statements about whether development should slow down. The panic is at an all-time high and the coverage writes itself.
Frontier firms linking arms to argue for limits on the thing they build is not new, and it is not the interesting part. The interesting part is what the noise is drowning out. This technology is already reshaping an older argument about the value of labor in our society, and the people standing closest to that change are the ones whose work has never been counted.
The headline is that AI is going to end the world. The trade-off is that fear isn't free, and the people paying for it are not the ones you would suspect.
Predicting the future is a fool's errand. Auditing the past is not.
Bill Gurley gave a speech this month arguing he cannot take AI doomerism at face value, and his reason was COVID. Seven years on, we still do not know the cause. Human and institutional failure produced a pandemic, and the after-action review that would tell us how never happened.
That is the cost of a decision-making culture that treats a conclusion as the end of the process. You run the cost-benefit analysis, you arrive at an answer, and the answer becomes a foregone thing that nobody audits.
My hometown is where they make fire trucks. A classmate walked me through the plant once, and the step that stayed with me was the one where they inspect how the truck got from idea to rolling off the lot. His point: repeating a mistake costs far more than taking the time to understand how the mistake got made.
Policy making almost never builds that step in. Which is why I ask the same question on this show every week. If we fail to audit the past, how do we price the future?
AI does not just accelerate technology. It accelerates norms.
It was our parents' generation who needed a husband to sign for a credit card or a bank account. That is one generation back, not a museum exhibit.
The policy choices that govern how women interact with the economy, whether a parent stays home to raise children or goes out for a paycheck, were made by people for whom one arrangement was simply the norm. Those choices are still load-bearing. The tax code assumes them. School enrollment assumes them. The workday that runs 9 to 5 against a school day that runs 8:30 to 3 assumes them.
Now add a technology that speeds up adoption. Not just adoption of tools, adoption of norms. If we accelerate into the norms we already have, we will not make invisible labor visible. We will make it cheaper to ignore. AI replacing that work does not make the work non-existent. It makes the record of it disappear.
The washing machine already ran this experiment
Ruth Schwartz Cowan wrote the book on this. More Work for Mother argued that the technological innovations that entered the home created real time savings for the person at home, who at the time was overwhelmingly a woman, and that the cost of those savings was a rise in expectations. Washing machines gave you time. In exchange, clothes were expected to be washed more often, and the standard climbed until the gain was gone.
Stanford sociologist Marianne Cooper updated the argument in TIME earlier this year with a sharper question: can AI free women from the mental load? That is the jump worth making. The physical workload of running a household is one thing. The invisible workload, the mental inventory required to keep a household running without failure, is another, and it is the part no device has touched.
The pedagogy here is well founded. Introducing technology into a system that has not priced the labor tends to raise expectations for the person implementing it rather than reduce their workload.
The first serious number, and why it is still too small
We have barely started measuring the physical load, let alone the mental one. But we started. Last week the Government Accountability Office estimated the national market value of unpaid household labor at close to six trillion dollars in 2024. That is roughly twenty percent of the entire economy.
Sit with that for a second, because a US government agency put a price on housework for the first time in any serious way.
Then notice what it leaves out. The estimate prices the work happening inside the home. It does not price the labor being replaced when a parent stays home. It counts the mother doing two loads of laundry a day. It does not count the value she contributes by watching her three children instead of buying childcare. She is doing two jobs, and the six trillion dollar figure accounts for one of them.
As a society we have not priced that, because for all of modernity it has simply been excluded as a cost all of us have been paying.
Zero times anything
Here is where the AI debate and the accounting problem collide.
If the mental load of household labor is valued at zero, and AI allows the person carrying it to hand off part of that labor, any number times zero is still zero. The productivity gain is real for the family and invisible to the economy. It shows up in no statistic, no hearing, no fiscal note.
And an economic contribution nobody records is one nobody has to pay for, staff for, or legislate around.
If this is an accounting question, then it has an answer. We can assign a value to what AI and new technology do for the economic trade-offs that a forgotten constituency is making every day. That constituency is working families.
The numbers that exist so far
PwC estimates that AI assistive support for unpaid household care and life management unlocks $330 billion in socioeconomic value, at a 40% adoption rate. Fewer than half of households, and the figure is still in the hundreds of billions.
A recent analysis of Japan and the UK found automation could cut three hours a day of unpaid work for women, fifty to sixty percent of the total. The consequence is the part to underline. That was enough for 9.3% of working-age women in Japan and 5.8% in the UK to take a paid job, purely by replacing unpaid labor at home.
Read that as a labor supply finding, because that is what it is. Every workforce model I have seen this year treats AI as a substitute for paid work. In these two markets it functioned as a release valve on unpaid work, and the workforce grew.
What I am not arguing
I am not arguing that the entire debate about new technology should collapse into one criterion about working families. I am arguing that we have dropped that empirical question entirely, in almost all of our policy making, at the exact moment it became critical.
Millennial parents are tired. They are burnt out. They are trying to enter the next stage of life while carrying societal dysfunction that got handed to them.
One promise of AI worth taking seriously is that a parent can run a household more smoothly without giving up an ambitious professional life. On net that is good for everyone. It is good for an employer whose employee is not fielding household logistics at their desk. It is good for children whose parents get quality time instead of wading through the minutia of raising kids in this particular society.
Paid family leave would be a good start. So would a school calendar that acknowledges both parents work. Families should get to choose the shape that works for them, and right now the choosing is mostly about mitigating exposure to a system that ignores the reality.
The ledger was built for somebody else
Millennials are the first cohort whose economic contribution does not fit the categories of the ledger built to track them. The blind spots have been treated as anomalies for an entire generation. They are the rule. We have been castigated as the generation that killed most everything, and I think it is time we were given our due for trying to build a society that makes room for the ones coming behind us.
If we can't audit the past, we can't price the future. I remain an optimist that the cost of the future can come down while the value of it has never been higher.
The Tradeoff publishes weekly. Ten minutes on the economic headline and what it actually costs you. Listen to this episode, and follow on Apple Podcasts or Spotify.