Is AI destroying jobs or creating them in 2026?

AI Created 1M Jobs, Cut 200K — What That Means for You

Neither panic nor complacency fits the September 2026 numbers: The Economist estimates AI has created roughly 1 million American jobs against about 200,000 AI-linked layoffs since mid-2023, and the New York Fed finds just 4% of service firms cut staff over AI. Davenport and Mittal's All-In On AI predicted the shape — returns flow to people who lead adoption, not those who brace for replacement.

On September 4th the Bureau of Labor Statistics reported that the American economy added 162,000 jobs in August — against a forecast of 65,000 — and unemployment held at 4.1%. That is not what an apocalypse looks like. I coach executives who have spent two years quietly bracing for AI to appear in the payroll numbers, and this week finally produced enough evidence to say something useful about where it actually appears.

The Economist pulled the threads together in a September 4th analysis and landed on a striking estimate: AI has so far created around 1 million new jobs in America, easily exceeding the roughly 200,000 layoffs attributed to AI since mid-2023. The displacement is real — Challenger, Gray & Christmas counts about 16,000 AI-related job cuts a month this year, and hiring in professional and business services is running about 10% below its 2015-19 average — but it drowns in a labor market that sheds roughly 1.7 million workers in a typical month anyway. Meanwhile the build-out absorbs people at speed. Data-centre construction is proceeding at more than $75 billion a year, nearly 60% higher than a year earlier; the five industries at the heart of the build-out have added roughly 320,000 more jobs since 2023 than broader trends would predict; and Indeed finds installation and maintenance roles at data centres advertising wages about 40% above comparable work. The white-collar side is moving too — LinkedIn data shows postings for heads of AI, AI engineers, and directors of AI roughly doubling since 2023-24.

The New York Fed's September survey release adds the inside-the-firm view, and it is stranger than either camp expects. In its region, 61% of service firms now use AI, up from 40% last year and 25% in 2024; manufacturers doubled to 51%. Yet only 4% of service firms report laying anyone off because of AI, retraining remains the primary workforce adjustment, and the median AI-adopting service firm has just 17% of its workers actually using the technology. Read that last number twice. Adoption is officially everywhere, and daily use is still a minority sport inside almost every company.

This is the pattern Thomas Davenport and Nitin Mittal documented in All-In On AI: fewer than 1% of large companies are genuinely AI-fueled, the binding constraint is human — leadership, culture, skills — and the smart ones upskill people who already understand the business rather than bidding for scarce outside talent. Three years on, the macro data is rhyming with their thesis. Firms are not swapping people for AI so much as hunting for people who can make AI useful, and paying a premium when they find them. The new roles — heads of AI, forward-deployed engineers, annotators judging model output — are coordination and judgment jobs, which is exactly what you would predict when the technology is capable and the organizations are not.

Here is the AI-exposure audit I run with coaching clients once a quarter. It takes about an hour:

List the five tasks that fill most of your week and mark, honestly, which current tools could do. Done when a colleague would agree with your marks.
Establish whether you sit in the 17% who use AI daily or the 83% who do not — and if you lead a team, count them too. Done when you have the real number, not the official one.
Pick one workflow where you become the person who deploys AI for others; the retraining budgets in that Fed data are looking for volunteers. Done when someone else uses something you shipped.
Reprice your external story: if your title describes automatable output rather than judgment, rewrite how you describe your work before the market does it for you. Done when your one-line introduction names the judgment, not the task.

The limits of the cheerful reading deserve naming. Aggregates hide individuals — if you are one of the 16,000 monthly cuts, a construction boom in Virginia is cold comfort, and a data-centre electrician role does not absorb a laid-off copywriter. The infrastructure engine runs on capital expenditure that could throttle abruptly. Survey talk of retraining is cheap until budgets confirm it. And one resilient year for young workers is not a structural verdict. What this week's data does kill is the excuse of waiting: the market is repricing toward AI-leading roles while most firms still have 83% of their staff on the sidelines. That gap is an opening, and openings close.

Sources: The Economist's September 4, 2026 analysis, "The jobs apocalypse is postponed. An AI jobs boom is here," read via Hindustan Times syndication (hindustantimes.com); Federal Reserve Bank of New York, Liberty Street Economics, "Businesses Are Using AI to Transform Work, Not Cut Jobs," September 2026 (libertystreeteconomics.newyorkfed.org); BNN Bloomberg's coverage of the August jobs report (bnnbloomberg.ca); Thomas H. Davenport and Nitin Mittal, All-In On AI (Harvard Business Review Press, 2023).