AI Adoption at New York Firms Jumped Again. The Share Reporting AI Layoffs Is 4%.

The Federal Reserve Bank of New York put special questions about artificial intelligence into its August regional business surveys and published the answers on September 1. The headline finding is that adoption has kept climbing steeply while the labour-market consequences firms report remain small — and that the consequence firms report most often is retraining, not dismissal.

The post, "Businesses Are Using AI to Transform Work, Not Cut Jobs," is by Jaison R. Abel, Richard Deitz, Natalia Emanuel and Nick Montalbano, and draws on the Empire State Manufacturing Survey and the Business Leaders Survey, which cover firms in New York State and northern New Jersey.

Adoption keeps rising

• Service firms using AI: 61% in 2026, up from 40% in 2025 and 25% in 2024.
• Manufacturers using AI: 51% in 2026, up from 26% in 2025 and 16% in 2024.

Both series have roughly doubled in two years. In the same period the share of manufacturers using AI has gone from one in six to about one in two.

But the adoption is shallow

The second set of numbers is the one that constrains how much the first can mean. About 75% of service firms and more than 90% of manufacturers describe their AI investment as minimal to modest. Roughly 15% of service firms say they have committed significant resources, and about 5% call it a major strategic investment. At the median adopting firm, 17% of workers at service firms and 7% at manufacturers actually use the tools.

A firm where seven workers in a hundred use a tool their employer describes as a modest investment is a firm that has adopted AI in the survey's sense. It is not a firm that has restructured around it. Read the 61% and 51% adoption figures with that in mind.

What it did to payrolls

Service firms using AI

• Laid workers off because of AI: 4%, up from 1% in 2025.
• Hired fewer workers than they otherwise would have: 15%, up from 12%.
• Hired more workers because of AI: 13%, similar to 2025.
• Retrained existing workers: more than 33%.

Manufacturers using AI

• Laid workers off because of AI: 0%, unchanged from 2025.
• Hired fewer workers than they otherwise would have: a small increase from 0% in 2025.
• Hired more workers because of AI: 0%.
• Retrained existing workers: more than 20%.

At service firms, then, retraining outnumbers AI-attributed layoffs by roughly eight to one, and the share hiring more because of AI, at 13%, is more than three times the share laying people off. The direction of travel is nonetheless upward on the layoff line: 1% to 4% is a small base but it is not nothing, and it is the number to watch when these questions are asked again.

AI has been more likely to augment workers than replace them.

Why the non-adopters matter

Roughly half of firms not using AI said their work is not suited to it. About a quarter said the technology is not yet capable enough. More than a third each cited data privacy and security concerns, accuracy and reliability concerns, and a lack of in-house technical skills. Cost was the least-cited barrier of all.

That last point is the useful one. If price were the binding constraint, adoption would be a function of the falling cost of inference, and the curve would keep bending upward on its own. Firms are instead describing constraints — task fit, reliability, trust, internal skills — that get resolved slowly and firm by firm, if at all. This is the strongest thing in the data against extrapolating the 2024-to-2026 adoption line forward in a straight line.

How this fits the earlier reading

The New York Fed's research director, Kartik B. Athreya, published a companion post on August 5 using the 2024 and 2025 waves, in which service-firm adoption ran 25% then 40%, with 44% expected within six months, and manufacturer adoption 16% then 26%, with 33% expected. Both realised 2026 figures came in well above what firms had projected for themselves. Firms have been underestimating their own adoption and, so far, not turning it into headcount reductions.