THE MACHINE AND THE VALLEY | PART I

At a courthouse window in Phoenix, a $69 filing and a conversation about Frankenstein led me toward a larger question: What if the next Industrial Revolution is already underway — and we are measuring it with the wrong numbers?

By Andrew Edling

The Maricopa Meridian

September 13, 2026

I met Anna through the window at Moon Valley Justice Court.

I had come to file paperwork. She handled the documents, I got my stamp, and I paid $69.

Nothing about the encounter suggested I was standing anywhere near the edge of a historic economic story. It was an ordinary trip to an Arizona courthouse: forms, fees, signatures, the routine administration of other people’s problems.

Anna was nice. She knew I wrote for the Maricopa Meridian, and we talked.

Her favorite book is Mary Shelley’s Frankenstein.

I told her I thought people misunderstood it. They remember the monster, but the book is much more psychological than that — ambition, loneliness, rejection, creation and the responsibility that follows when somebody brings something powerful into the world without fully understanding what comes next.

Before I left, Anna wrote something on a Post-it note for me.

Her Starbucks order.

Pink Drink. Coconut milk.

Then she told me something considerably harder to forget.

This area, she said, was No. 2 in the country for evictions.

I went home and checked.

She was substantially right.

Princeton University’s Eviction Lab defines its Phoenix tracking area as Maricopa County. As of its August update, it recorded 84,511 eviction filings during the preceding 12 months, a filing rate of about 13 cases for every 100 renter households. Among the city areas in Eviction Lab’s current tracking system, Phoenix’s raw total trails only New York City; it is higher than Houston, Las Vegas, Memphis and every other city area in that dataset.

The previous year was worse. Eviction Lab says landlords filed a record 86,946 eviction cases in the Phoenix area during 2024 — roughly one every six minutes.

And the cases have not stopped arriving. Maricopa County Justice Courts recorded 7,743 eviction filings in August 2026.

An eviction filing is not the same thing as an eviction. It is a lawsuit. Some cases are dismissed. Some tenants pay. Some settle. Some people leave before a constable ever reaches the door. Court records alone cannot tell us how many of those 84,511 filings ended with somebody physically losing a home.

But 84,511 lawsuits are still telling us something.

They are telling us that tens of thousands of times in a single year, the relationship between household income and housing expense deteriorated far enough to reach a courthouse.

And as I began looking at those numbers, I found another set of numbers that seemed, at first, to belong to an entirely different story.

They were about jobs.

Phoenix built an economy around the office

In May 2021, according to the Bureau of Labor Statistics, the Phoenix metropolitan area employed 92,970 customer-service representatives.

That was extraordinary.

Phoenix had the second-highest raw number of customer-service representatives among American metropolitan areas and, among large metros, one of the greatest concentrations anywhere in the country. Customer-service jobs were more than twice as concentrated here as they were nationally. Phoenix also employed about 326,500 people across office and administrative support occupations.

Four years later, the picture looked different.

By May 2025, Phoenix employed 68,930 customer-service representatives.

That is roughly 24,000 fewer jobs — a decline of about 26 percent between the two BLS snapshots.

Overall office and administrative support employment had fallen from about 326,500 to 295,910, a reduction of more than 30,000 positions.

Those numbers do not prove artificial intelligence eliminated those jobs. The period contains the aftermath of a pandemic, changes in remote work, outsourcing, business cycles, ordinary automation and changes in the composition of the regional economy.

But then, in August 2026, the Bureau of Labor Statistics published something that made the Phoenix numbers considerably more interesting.

For the first time, BLS classified the relative artificial-intelligence exposure of all 831 detailed occupations for which it publishes employment projections.

Customer-service representative was placed in the “very high” AI-exposure category. BLS says that designation means a comparatively large portion of the job’s tasks can potentially be assisted or completed by artificial intelligence and that large language models have already been observed performing some of those tasks. BLS is careful to say exposure is not a prediction that a job will disappear.

Then the agency went further.

BLS now projects that American employment of customer-service representatives will decline by about 142,000 jobs between 2025 and 2035. In explaining the forces affecting that occupation, BLS specifically cites advances in speech and language processing and artificial intelligence that allow computers to handle customer interactions more frequently, reducing the number of human representatives required.

It is not just customer service.

BLS projects office and administrative support employment overall to fall 4 percent during the coming decade, shedding roughly 752,100 jobs — the largest numerical decline of any major occupational group. General office clerks, customer-service representatives, secretaries, bookkeeping clerks and several other familiar office occupations are all projected to shrink.

That matters in Phoenix because Phoenix built a remarkable amount of its modern economy around precisely this kind of work.

The cubicle. The headset. The claims department. The collections desk. The billing office. The call center. The back office.

For decades, these were not glamorous jobs. That was partly their importance.

A person did not need to invent a company or graduate from Stanford to get one. A high-school diploma, a phone voice, training and reliability could open a door into the middle of the economy.

Now the economics of that door are changing.

An Industrial Revolution can begin without mass layoffs

Our cultural image of automation is almost useless.

We imagine a robot replacing a worker. One machine rolls in. One human walks out.

But software can change employment without creating a dramatic firing scene.

Suppose 10 people once performed a body of work. Artificial intelligence makes each of them 20 percent more productive. The company does not have to fire two people tomorrow. Two people may retire. One may resign. Another may transfer. The company simply does not replace all of them.

The work remains. The positions disappear.

That distinction may explain why one of the clearest early signals of AI’s labor impact is appearing not in layoffs, but in hiring.

Stanford’s Digital Economy Lab examined payroll data covering millions of American workers through June 2026. Its researchers found no widespread, economy-wide displacement attributable to AI.

But they found something striking among workers ages 22 to 25.

Employment among young people in occupations highly exposed to artificial intelligence was about 19 percent below where it would have been had those workers kept pace with similarly aged workers in less-exposed occupations.

Experienced workers did not show the same gap.

Most importantly, Stanford found that the adjustment appeared to be happening primarily through reduced hiring of young workers rather than increased firing of existing ones.

That may be the sentence future economic historians care about.

Because an economy can change profoundly before unemployment explodes.

The first casualty of a technological revolution may not be the worker who gets fired. It may be the worker who never gets hired.

The junior developer. The new paralegal. The first-year analyst. The entry-level copywriter. The customer-service trainee. The assistant whose old job taught the skills required to someday become the manager.

If artificial intelligence allows experienced workers to produce more, businesses can benefit tremendously. But the same productivity gain creates a question that has received far less attention:

What happens to the bottom rung of the ladder?

The machine is already inside the building

This is not a future technology awaiting deployment.

The Federal Reserve reported this year that one in four American workers had used generative AI as part of their job during the previous month. Among those users, 81 percent said AI saved them time. Fifty-two percent said it improved the quality of their work. Fifty-five percent said it enabled them to perform new tasks.

The Census Bureau approached the question from the company side.

During late 2025 and early 2026, 18 percent of American firms reported using artificial intelligence in at least one business function. Because adoption is greater at larger employers, those firms represented roughly 32 percent of American employment.

Among very large companies in information, professional services and finance, employment-weighted adoption reached roughly 60 to 70 percent.

Stanford’s 2026 AI Index provides the historical comparison.

Generative AI reached approximately 53 percent adoption within three years, faster than either the personal computer or the internet reached comparable adoption after their mass-market introductions.

Think about what we are saying.

The personal computer reorganized the office. The internet reorganized commerce, communication, publishing, media, retail and eventually nearly everything else.

Generative artificial intelligence is diffusing faster.

And somehow much of America continues to discuss it as though it were an interesting new application on a phone.

The layoffs are beginning to carry a name

Some employers are no longer subtle about it.

Through August, the outplacement firm Challenger, Gray & Christmas says American employers cited artificial intelligence in 116,175 announced job cuts during 2026, about 22 percent of all announced cuts. AI had been the leading reason employers cited for job cuts for five consecutive months before August.

There are important limits to that number. Challenger records employers’ stated reasons for announced cuts; it is not independently proving that an algorithm caused each termination. And employers are still hiring. In August, the United States added 162,000 payroll jobs, while unemployment remained at 4.1 percent.

This is not an AI depression.

That is precisely why the moment is so interesting.

If unemployment were 12 percent and millions of workers had already been displaced, there would be no investigative mystery left. Congress would be holding hearings. Presidential candidates would have talking points. Economists would have named the crisis.

Instead, the headline labor market still looks functional. Underneath it, the composition of work is changing.

Stanford sees the effect first among younger workers. BLS sees the occupations whose task structure is most exposed. Census sees the technology moving into firms. The Federal Reserve sees workers already using it. Employers themselves are beginning to attach the letters A.I. to job cuts.

The transformation is visible. It simply has not yet arrived in the form Americans have been trained to recognize as a crisis.

Arizona has less room for error than the national averages suggest

The latest Arizona employment report adds another layer.

The state unemployment rate stood at 4.9 percent in July, above the national rate. More unusually, Arizona’s seasonally adjusted labor force was 92,368 people smaller than a year earlier, a decline of 2.4 percent, even while nonfarm employment remained modestly higher.

Again, that decline cannot simply be labeled “AI.”

But it describes the landscape onto which AI is arriving.

Phoenix has hundreds of thousands of workers in office and administrative occupations. The city has historically had an exceptional concentration of customer-service employment. Those jobs have already declined sharply from their 2021 level.

The federal government now identifies customer service as among the occupations with the greatest relative AI exposure and explicitly predicts that technological productivity will reduce the number of humans required for the work.

And in the same metropolitan county, landlords filed more than 84,000 eviction cases during the most recent 12-month period tracked by Princeton.

These are separate datasets. They should not be dishonestly welded into a causal claim.

No public record I found establishes that artificial intelligence caused Maricopa County’s eviction filings.

What the records establish is something different:

Phoenix is entering a historic transformation of work with a large population of renters already demonstrating extraordinary financial vulnerability.

That is not causation. It is exposure.

Follow the money

There is another way to judge whether something resembling an Industrial Revolution is underway.

Watch what capital does.

The Bank for International Settlements — an institution whose members include the world’s central banks — says the five largest technology companies alone are set to spend more than $1 trillion on AI-related capital expenditure during 2025 and 2026. Industry estimates cited by the BIS suggest global AI investment could reach $3 trillion to $4 trillion by 2030. Increasing portions of that expansion are being financed through debt and private credit.

Then the BIS did something remarkable. It reached for history.

Canal mania in the 1830s. British railway mania in the 1840s. Electrification in the 1920s. The dot-com surge in the 1990s.

Each boom formed around a technology that really did change the world. Each also attracted periods when capital ran ahead of immediate economic returns. The BIS warns that AI investment could likewise produce financial instability if expected returns fail to arrive quickly enough, while explicitly stopping short of predicting that such a bust must occur.

There is something important in that comparison.

Transformative technologies and speculative excess are not opposites.

Railroads really changed civilization; railway investors still lost fortunes. The internet really transformed the planet; the dot-com bubble still burst.

Artificial intelligence can be both profoundly useful and economically disruptive. Those ideas can coexist.

We keep looking for the robot

Maybe that is why we are having trouble seeing what is happening.

The first Industrial Revolution left evidence everywhere: coal smoke, factories, rail lines, machine noise, cities physically transformed by industry.

This revolution is quieter.

Its factory may be a data center in another state. Its machinery may appear to an employee as a chat box inside software the company already uses. Its first economic effect may be an email saying a hiring requisition has been withdrawn.

The worker whose job changes may never see the machine responsible. There may be no machine to see. Only software.

That makes the transition psychologically strange.

A person knows when a factory closes.

What does a 23-year-old know when the job he expected simply never appears? What does the college graduate blame after sending 300 applications? What does a customer-service representative think when four people leave the department and management replaces only two? What does a manager call the savings when the same staff suddenly completes 30 percent more work?

Productivity. Efficiency. Transformation. Automation. Artificial intelligence.

Those can all describe the same event depending upon which side of the payroll system you occupy.

I keep coming back to Frankenstein

I keep thinking about Anna.

About the courthouse. The $69. The Pink Drink written on the Post-it note. And Frankenstein.

Mary Shelley published the novel in 1818, during the era in which industrialization was beginning to transform Britain and, eventually, the world.

I do not think artificial intelligence is Frankenstein’s monster. That metaphor is too obvious and ultimately wrong. The technology is not a creature stalking its creator. It is a tool, an infrastructure, an economic force built and deployed by people and institutions making choices.

But Shelley understood something that remains relevant two centuries later.

Creation is only the first question. Responsibility comes afterward.

If artificial intelligence makes an employee twice as productive, who receives the benefit? If companies require fewer junior workers, where will tomorrow’s senior workers learn? If hundreds of thousands of clerical jobs disappear gradually through attrition rather than mass layoffs, what system notices? If productivity rises while the number of people required to produce that output falls, how quickly can labor move into whatever comes next?

And what happens during the years between the old economy and the new one?

Those are not arguments against artificial intelligence. They are questions about an economic transition.

Every major technology creates winners. The great ones create enormous amounts of wealth.

They also destroy arrangements that once seemed permanent.

History will put a date on this

One day, textbooks will explain what artificial intelligence did to the early 21st-century economy. They will have charts. They will have neat periods: Before AI. The transition. After AI.

Historians will argue over the year the revolution truly began: 2022, when ChatGPT appeared; 2024, when corporate investment exploded; 2025, when workplace adoption became widespread; 2026, when the labor data began showing an unusual fracture among young workers.

History has the luxury of knowing what happened next. We do not.

That is the strange privilege of living inside the event.

The evidence today does not show mass technological unemployment. Stanford says so directly. The United States added jobs last month. Many occupations exposed to AI will survive, change or even grow. AI may generate industries and professions we have not yet imagined.

But absence of catastrophe is not absence of revolution.

The public record already shows a technology spreading faster than the PC and internet, entering workplaces representing a large share of American employment, increasing the productivity of the humans who use it, altering hiring patterns among young workers, influencing federal employment forecasts and attracting trillions of dollars of capital.

Here in Phoenix, another public record shows something more local.

A city that once employed nearly 93,000 customer-service representatives now employs fewer than 69,000. A federal agency says that occupation is among those most exposed to AI and expects technology to reduce future demand for the workers performing it. A state labor force has contracted.

And in Maricopa County, more than 84,000 eviction filings crossed courthouse counters during the past year.

We cannot yet draw a straight line from the machine to the courthouse.

But we would be fools not to watch whether one develops.

I had gone to Moon Valley Justice Court expecting to leave with a stamped document. Instead, a clerk named Anna told me something about evictions that made me start pulling on a thread.

I am still pulling.

Maybe years from now the economic historians will decide that this was the period when artificial intelligence stopped being a technology story and became a labor story. Maybe they will conclude the transformation produced abundance on a scale we can barely imagine. Maybe they will write about the workers and institutions that failed to adapt quickly enough.

Probably they will write about both.

The first Industrial Revolution announced itself with steam, iron and smoke. This one came through a browser window.

And by the time everyone agrees that the revolution happened, it may already be history.

EDITOR’S NOTE

This article is the first installment of an ongoing Maricopa Meridian enterprise series examining how artificial intelligence is changing the Valley’s labor market — and whether those changes are beginning to appear elsewhere in Maricopa County’s economy.

The Meridian is reviewing public employment data, court records and individual eviction cases to follow what happens when workers lose hours, positions or access to traditional entry-level employment.

This first report establishes what the public record already shows: Phoenix has lost tens of thousands of customer-service and administrative jobs from recent peaks; artificial intelligence is rapidly entering precisely those categories of work; young workers in highly exposed occupations are experiencing unusual hiring weakness; and Maricopa County simultaneously carries one of the heaviest eviction filing volumes among major metropolitan areas tracked nationally.

Those facts do not establish that artificial intelligence is causing Maricopa County evictions.

That is the question our reporting will test.

Coming next: Part II: From the Cubicle to the Courthouse

Reporting note: Eviction figures measure court filings, not completed removals or unique households displaced. Occupational employment changes between BLS survey years cannot by themselves establish why jobs increased or disappeared. AI exposure classifications are not predictions that individual jobs will be eliminated. No evidence reviewed for this article establishes that artificial intelligence caused Maricopa County’s eviction volume; the article examines the documented technological transformation of Phoenix’s labor market alongside the county’s existing housing vulnerability.

Disclosure: Andrew Edling owns The Wolf Legal Services, an Arizona process-serving and investigations business. No confidential client information or restricted investigative data were used in this report.