John Rector / The economy in the Age of AI
The Hiring Signal Has Changed
A job posting is not a hire. Start with who actually joins a payroll, then ask how AI changes what that means for interest rates, inflation, and stocks.
JOLTS · rounded estimate
JOLTS · share of employment
BEA · September 30 release
For someone trying to start a career, a job advertisement is a possibility. Joining a payroll is an outcome. Our reading of the AI economy should begin with that distinction.
A company can grow without hiring proportionately more people. That makes completed hiring—and who can access it—essential to understanding what AI changes.
The familiar market shorthand runs like this: hiring cools, household demand weakens, inflation retreats, the Federal Reserve cuts interest rates, and stocks benefit from cheaper money. Even historically, that chain was conditional. A recession could damage earnings faster than falling rates helped valuations.
AI introduces another complication. The company slowing its recruiting may still be expanding production. It may be redirecting spending from additional employees toward computing, software, and the people needed to make those systems work. The hiring number tells us something important. It no longer tells us enough.
Count the people who actually got hired
My 24-year-old students tell me they are saturated with what they call “ghost jobs”: listings they no longer trust to lead to an actual hire. They have lost faith in the language of “job creation.” Their practical question is more concrete: Who actually got onto a payroll?
That classroom experience deserves to shape the question. It does not establish how many advertised positions are fake, or mean that every unanswered application went to a ghost listing. A genuine vacancy can remain unfilled. But an advertisement is plainly not evidence that someone started work.
Yes, actual hires are measured. The Bureau of Labor Statistics’ Job Openings and Labor Turnover Survey, or JOLTS, estimates additions to employers’ payrolls during the month. It includes new hires, rehires, seasonal and part-time workers, and certain transfers between locations. These are hiring events, not a count of unique people landing their first career job. [9]
August’s seasonally adjusted estimate was 5.192 million hires across the nonfarm economy, including 4.846 million in the private sector. The total hires rate was 3.3% of employment. Private-sector hires were almost identical to July’s 4.847 million and August 2025’s 4.851 million. BLS described total hiring as little changed over both the month and year. These are survey estimates subject to revision, not an individually audited register of every start. [3]
That evidence requires a more careful claim: the latest national actual-hires measure does not establish a fresh hiring collapse. The ICIMS platform reported U.S. hires falling 3% in July and 1% in August, while remaining 2% above its year-earlier baseline. That describes its measured recruiting activity; it cannot override the national series or stand in for every employer. [2]
Three measures that should never be interchangeable
A posting is an invitation to apply. Even the official JOLTS openings measure is more specific than a tally of online ads: it requires an available position, the possibility of starting within 30 days, and active external recruiting. It still measures an unfilled position, not a completed hire. [9]
A hire is an addition to a payroll. It can fill a newly created role or replace someone who left. For a person seeking work, both can be meaningful opportunities. Millions of hires can happen without much net employment growth because workers are also leaving.
Net payroll growth is the change in the number of paid jobs. The familiar monthly “jobs added” headline comes from the establishment survey, which measures payroll employment for a reference pay period. It does not count job advertisements. Ghost listings therefore do not directly inflate that figure. It is a different measure from the full month’s gross hires, using a different survey and reference period. [10]
The October 2 report estimated September net payroll growth of 29,000 and unemployment of 4.2%. Those remain useful macroeconomic observations. They are not 29,000 advertised openings, nor the total number of people hired that month. [1]
Neither a national payroll gain nor millions of gross hires establish that entry-level professional work is accessible to these students. The August JOLTS release breaks hiring out by industry and establishment size, not by age or first-career-job status. Its total cannot answer their specific question. [3]
My priority for judging access is completed hires, then the quality and durability of those jobs. To understand the economy, I would read that evidence alongside separations, employment, hours, and income. Interest costs, uncertainty, and labor availability can all restrain hiring; the data do not isolate AI as the cause.
The other side of hiring: who is leaving?
My students also need to understand the retirement wave. A company does not have to expand its headcount to hire them. It may need someone because an experienced employee is leaving. In late 2026 and beyond, baby-boomer departures belong beside AI in any serious discussion of access to work.
Boomers were born from 1946 through 1964. By the end of 2026, they will be 62 to 80; the youngest turn 65 in 2029. That is a demographic timetable, not a retirement schedule. [11]
The Alliance for Lifetime Income’s January 2025 research announcement projected 4.18 million Americans turning 65 that year—about 11,400 a day—and described 2024–27 as a peak period. Those are birthdays, not observed workforce departures. Some people retired earlier; others continue working. BLS reports that 19.1% of Americans age 65 and older were working or looking for work in 2025. [12] [13]
How many boomers are actually leaving? A historical benchmark is Pew’s finding that the boomer labor force shrank by an average of 2.2 million people annually from 2010 through 2018. That is a net decline in a birth cohort, not a count of gross retirements, and it must not be presented as the 2026 rate. The current releases cited here do not supply a comparable boomer-only annual exit figure. [14]
Looking ahead / BLS projections, 2025–35
7.36 million projected labor-force exits per year, across all ages and reasons.
9.51 million projected transfers to different occupations per year.
17.46 million projected occupational openings per year, including the above separations and roughly 592,000 from annual net employment growth.
Rounded from BLS Table 1.10. These are modeled annual averages, not observed hires, online listings, or boomer-only retirements. [15]
BLS’s exit category includes retirement and departures for other reasons at any age. Occupational transfers mean moving into different kinds of work; they exclude simply changing employers in the same occupation. These projections therefore cannot be directly compared with twelve months of JOLTS hires. [16]
A simple hypothetical makes the opportunity concrete: a company has 100 employees, ten retire, and it hires ten replacements. It still has 100 employees. Net job growth is zero. Ten people nevertheless got hired. Retirement may also trigger a chain of internal promotions that eventually opens an entry-level position.
The AI twist is that the replacement is no longer assured. An employer might refill the role, combine its responsibilities with another job, or automate some tasks and hire someone with different skills. Retirements could let a business reduce staffing through attrition instead of layoffs. That is a possible adjustment mechanism, not proof of how employers are handling every departure.
For students, the useful question becomes: Which employers are actually replacing departing workers, and what work survives the redesign? Ask about recent replacement hires, training, succession plans, and the responsibilities a new employee will inherit. Demographics creates a reason to investigate; a completed hire remains the evidence.
The macroeconomic implications run both ways. Departures can constrain labor supply and sustain wage pressure even when net employment barely grows. AI could offset some lost capacity. Retirees also continue consuming, with spending supported by pensions, benefits, and savings. Neither inflation relief nor lower interest rates follows automatically. For businesses, lost experience and training costs matter alongside any payroll savings.
Watch the job that never gets posted
Imagine a company whose existing team can handle more customer requests with AI assistance. Instead of filling every departure or approving the next hiring round, it redesigns the work. No dramatic layoff announcement is necessary. The change appears first as fewer openings for newcomers.
That is a plausible mechanism, not an established explanation for the entire slowdown. Governor Michael Barr said on September 29 that evidence of significant economy-wide displacement remained limited, even as some entry-level opportunities in exposed sectors appeared affected. [4]
The distinction matters for workers. An incumbent might become more productive while someone trying to enter the profession finds fewer doors open. Aggregate resilience and individual difficulty can coexist.
Interest rates: weak hiring is no automatic invitation to cut
The current backdrop already challenges the old reflex. On September 16, the Fed raised its policy range by a quarter percentage point to 3.75%–4%, citing elevated inflation alongside solid activity, strong productivity, and robust capital investment. [5]
My reading is that September’s weaker jobs report increases the reason for caution about further tightening. It does not, by itself, establish the case for a cut. The diagnosis still matters: insufficient demand, greater productive capacity, and a mismatch between workers and available jobs require different responses.
Cheaper borrowing can support demand. It cannot directly turn a displaced analyst into an experienced electrical contractor or recreate a task a company has automated. Stimulus applied to a skills mismatch can increase spending without quickly fixing employment.
There is also a less intuitive possibility: successful AI could support higher long-run real interest rates. Barr described how stronger expected investment returns and changes in saving could raise the equilibrium rate—the rate compatible with balanced economic activity. He stressed that it was too early to know whether that was happening. [4]
And the Fed’s overnight rate is not the ten-year Treasury yield. Longer yields also reflect expected inflation, future short rates, and compensation for holding long-term debt. A policy cut would not guarantee an equal decline in mortgage or corporate borrowing costs.
Inflation: the buildout can arrive before the savings
August PCE inflation was 3.4% over the preceding year, or 3.0% excluding food and energy. Real consumer spending increased 0.6% that month. Those September 30 BEA figures describe an economy with persistent inflation and continuing demand, despite the weak hiring backdrop. [6]
AI can eventually reduce the labor required per unit of output. That creates room for better margins, higher wages, lower prices, or some combination. Competition and bargaining determine how much of the benefit customers and workers receive.
Meanwhile, building the capacity requires physical inputs. Governor Lisa Cook’s September 28 remarks identified potential pressure from AI investment on shared resources such as energy and construction labor. She expected modest productivity-related disinflation over the next few years, while emphasizing uncertainty about its timing and breadth. [7]
The analytical implication is a timing problem: investment demand can arrive before widespread efficiency gains. Cooler office hiring could coexist with expensive power, constrained construction capacity, and stubborn inflation. AI makes some tasks cheaper; it does not instantly make every input abundant.
Stocks: the economy and the index can tell different stories
A company that maintains sales while reducing the labor needed to deliver them may improve operating margins. Shareholders can benefit even while the hiring environment becomes less welcoming. A stock index concentrated in successful firms can therefore rise while many job seekers struggle.
That is a scenario, not a prediction that weak employment is bullish. AI spending carries costs: equipment depreciation, energy, integration, oversight, and continuing model use. Revenue per employee can rise because of inflation, outsourcing, or layoffs; it is not sufficient proof of better productivity.
The useful questions concern cash flow after investment, customer retention, durable competitive advantage, and whether the gains are already reflected in valuations. A technology can transform an industry while disappointing investors who paid for even more extraordinary results.
There is a wider feedback loop, too. Payrolls fund customers. If lost labor income outruns new opportunities, household spending can weaken and eventually hurt the same businesses celebrating efficiency. The distribution of AI’s gains is part of the earnings outlook.
Conditional scenarios / not forecasts
One hiring signal. Three possible economies.
These mechanisms can overlap. The task is to determine which is becoming more important.
Demand is fading
Sales soften, hours fall, and companies need fewer workers.
Rates: A stronger case for easing if inflation permits.
Stocks: Lower yields compete with weaker earnings.
Inflation: Less spending tends to reduce pressure.
Productivity is spreading
Real output grows faster than hours worked, with efficiency gains reaching more firms.
Rates: No automatic decline; investment demand matters.
Stocks: Earnings may improve, subject to costs and valuation.
Inflation: Relief depends on supply growth and price pass-through.
Reallocation is painful
Some roles shrink while demand for different skills and scarce inputs expands.
Rates: Broad stimulus may struggle to fix the mismatch.
Stocks: Larger differences between beneficiaries and disrupted firms.
Inflation: Bottlenecks can persist alongside employment weakness.
My prediction: Warsh will hold the line, with an emergency exception
My base case for the next six to twelve months is that Kevin Warsh will favor price stability and independent market signals over making money cheaper to satisfy politicians or support stock prices. My confidence is moderate. I expect him to tolerate uncomfortable bond yields and greater repricing rather than promise a comforting rate path, while preserving the option to intervene if financial markets stop functioning.
The core question is whether he will let the bond market speak for itself. The Fed deliberately sets a short-term policy rate; there is no untouched “true rate” it can simply announce. Longer-term market rates also include risk premiums and changing expectations. Here, “let the rate speak the truth” means allowing economic evidence to discipline policy and allowing markets to price risk. “Manipulation” would mean subordinating those goals to a preferred political or asset-price outcome. A rate change alone proves neither.
Warsh’s August 28 Jackson Hole speech supports the first interpretation. He argued for less routine forward guidance, clearer market signals, a firm 2% inflation objective, and sparing use of unconventional tools outside genuine crises. He also explicitly recognized that the Fed determines short-term rates. His specific concern is a feedback loop: markets trade on Fed guidance, then the Fed treats those market prices as independent information. He wants investors to assess economic conditions themselves. Applied to bonds, that means giving yields more room to express investors’ judgments, rather than steering them toward a preannounced path. It does not mean that every yield is correct or that the Fed relinquishes its policy rate. [8]
The September rate increase is more persuasive to me than a speech. It is an observable action consistent with putting inflation control ahead of immediate relief in borrowing costs. But it was a unanimous committee decision, not proof of Warsh’s private motives or a guarantee of future independence. [5]
The AI test will be whether he demands evidence before spending the productivity dividend. Lower rates could be justified if durable efficiency gains reduce inflation pressure. Cutting because those gains are merely promised would be a more speculative wager. My prediction is that persistent inflation will outweigh the AI promise in his near-term decisions.
I would revise that view if the Fed repeatedly eased while underlying inflation stayed elevated, employment remained resilient, and its explanation failed to establish a mandate-based reason. One cut, one presidential demand, or one falling stock index would not establish political capture. Conversely, holding firm through market criticism would strengthen the case for independence.
The exception is a breakdown in credit-market functioning. I expect him to distinguish emergency liquidity support from protecting investors against ordinary losses. Whether any intervention is narrow, temporary, and withdrawn after functioning returns will matter more than the fact that intervention occurred.
My call: he is more likely to let the bond market speak than to steer it toward a politically convenient answer. The harder test will arrive when honoring that discipline becomes costly. Judge him by the evidence he responds to, the risks he accepts, and the conditions under which he changes course.
The dashboard needs more than payrolls
Watch demand: real sales, consumer spending, hours worked, and unemployment claims.
Watch production: real output per hour and unit labor costs, across several quarters.
Watch access: actual hires and the hires rate by industry, supplemented by evidence on entry-level starts, unemployment duration, job retention, and movement into new occupations. Do not substitute posting counts for completed hires.
Watch the payoff: cash flow after capital spending, price reductions, wages, and whether gains spread beyond a few firms.
No single indicator isolates AI. Together, these measures help distinguish an economy running out of customers from one learning to produce differently—and reveal when both are happening at once.
In the Age of AI, we are not playing by exactly the same rules. The relationships between headcount, output, profits, and purchasing power can change. Scarcity, demand, and valuation still matter. What changes is how reliably one familiar number reveals the others.
Start with who actually got hired. Then ask what that work pays, how long it lasts, and who is being left outside. Only then can the hiring signal tell us something useful about the Age of AI.
Evidence and reading notes
- BLS: Employment Situation, September 2026 · October 2. Preliminary payroll data and revisions.
- ICIMS: September Workforce Report · September 18. Recruiting-platform observations; distinct from national statistics.
- BLS: Job Openings and Labor Turnover, August 2026 · September 29.
- Michael Barr: Economic Conditions and Monetary Policy · September 29. Individual policymaker’s assessment.
- Federal Reserve: September 16 FOMC statement.
- BEA: Personal Income and Outlays, August 2026 · September 30. Inflation figures use this release, which postdates the speeches’ earlier estimates.
- Lisa Cook: An Update on AI and the Economy · September 28. Individual policymaker’s assessment.
- Kevin Warsh: In Our Time · August 28, Jackson Hole. Primary source for his stated policy principles; the prediction is the author’s judgment as of October 3.
- BLS: JOLTS Data Definitions · Hires and qualifying openings.
- BLS: Current Employment Statistics Frequently Asked Questions · Paid employment, reference periods, and estimation.
- Pew Research Center: The oldest Baby Boomers turn 80 in 2026 · January 9, 2026. Cohort definition and demographic context.
- Alliance for Lifetime Income: Peak 65 research announcement · January 28, 2025. Industry organization’s demographic projection, not measured retirements.
- BLS: Nearly one in five older Americans in the labor force in 2025 · May 28, 2026. Observed annual participation rate.
- Pew Research Center: Baby Boomers are in the workforce later in life than past generations · July 24, 2019. Historical cohort decline; not a current exit rate.
- BLS Table 1.10: Occupational separations and openings, projected 2025–35 · Annual averages, in thousands; all occupations and ages.
- BLS: Employment Projections Data Definitions · Distinguishes exits, transfers, and modeled openings.
The mechanisms and scenarios above are the author’s analysis; the cited data do not establish that AI caused the aggregate hiring slowdown.
John Rector · Understand the change behind the number.