News · 2026-11-02

New York Fed analysis links small-business AI use to hiring expectations

Small businesses using AI reported stronger expectations of employment growth than non-users in an October 8 New York Fed analysis. The finding comes from a survey of business owners, not a count of jobs created after adopting AI. It offers evidence against a simple story of universal hiring pessimism among AI users, while leaving actual employment effects unresolved.

Key facts

• Will Aarons and Asani Sarkar published the analysis on October 8, 2026.
• The underlying 2025 survey’s AI module covered 5,248 employer firms.
• Net expectations of employment growth were 33 percentage points for AI users and 15 for non-users.
• The primary source is the authors’ Liberty Street Economics post.

The headline distinction is between a plan and a payroll. A business owner can expect to employ more people next year without having hired anyone yet. Aarons and Sarkar analyze that forward-looking expectation using the Small Business Credit Survey, fielded between September and November 2025. The responses were weighted to better match the national population of employer firms.

The most vivid number also requires a careful translation. The 33-point figure is a balance: the percentage of AI users expecting employment to rise minus the percentage expecting it to fall. The corresponding balance for non-users is 15 points. It does not mean AI-using businesses expect to add 33 percent more workers. It is a measure of how many more respondents lean toward expansion than contraction.

Think of it as a weather-outlook board. If more shop owners put their marker under expansion than under contraction, the board shows optimism about staffing. It does not tell you the size of every planned hire, whether plans will survive the next year or why the owners differ. Confusing that board with a headcount report turns a useful survey into a claim it cannot support.

The authors go beyond the raw comparison by accounting for firm and owner characteristics, location and prior employment changes. The association between AI use and more positive employment expectations remains. That makes the result more informative than simply observing that some AI adopters feel optimistic. It still does not create the randomized comparison needed to say adopting AI caused the change in expectations or will cause hiring.

Several alternative explanations remain possible in an observational survey. Firms with stronger growth prospects might both adopt technology and plan to hire. Managers who are more willing to experiment might make different forecasts. Statistical adjustment can address measured differences, but it cannot guarantee removal of every relevant unmeasured one. Ground Truth’s causal-graph lesson explains why choosing controls is itself a substantive decision.

The survey’s other responses help keep the narrative grounded. Most respondents reported no change in current labor costs from AI use, while a smaller share reported increased sales. Those findings can coexist with positive hiring expectations: an owner may believe a tool creates room to expand without observing an immediate reduction in wage spending. They do not establish that a productivity gain has already paid for new employees.

This is also not a clean referendum on every labor-market concern raised by AI. Small employer firms differ from large companies, and an aggregate expectation says little about entry-level roles or particular occupations. A firm could plan to grow overall while changing which workers it wants. Conversely, an individual job loss does not establish that all adopters expect to shrink. The useful question is which outcome a source actually measures.

Aarons and Sarkar explicitly reserve the outcome question. Whether adopters achieve the stronger results they expect “is a subject for future research,” they write. That short caveat prevents the tempting upgrade from optimism to observed job creation. Their post is a research contribution hosted by the New York Fed, not a guarantee about the next year’s hiring decisions.

For readers tracking AI adoption, the finding provides a more textured account than either mass displacement or effortless job creation. Smaller businesses may use tools to address operational limits and see expansion as more feasible. That is a plausible interpretation discussed by the authors; it is not a directly tested mechanism showing which tool removed which constraint. Ground Truth’s randomized-trial lesson explains the stronger design required to estimate a causal effect.

The story resurfaced in the November 2 intelligence cycle, but the analysis itself is from October and the survey responses from 2025. Those dates belong beside the finding. The next decisive evidence would connect adoption to subsequent payroll and revenue outcomes for comparable firms. Until then, the verified result is a persistent association with stronger expectations, not proof that AI has already increased small-business employment.

Primary source, verified: read the paper →

Ground Truth. (2026, November 2). New York Fed analysis links small-business AI use to hiring expectations. Ground Truth. https://groundtruth.day/news/new-york-fed-ai-adoption-hiring-expectations.html

@misc{groundtruth:new-york-fed-ai-adoption-hiring-expectations,
  title  = {New York Fed analysis links small-business AI use to hiring expectations},
  author = {{Ground Truth}},
  year   = {2026},
  month  = {nov},
  url    = {https://groundtruth.day/news/new-york-fed-ai-adoption-hiring-expectations.html}
}

Topics: economy · employment · small-business · ai-adoption · surveys