Small businesses using AI regularly across multiple functions are cutting outside services while reporting stronger hiring plans, a survey from U.S. bank Citizens found.
Among the group, almost all (94%) said they had reduced or eliminated spending on at least one external service.
Citizens surveyed 500 U.S. business decision-makers from Sept. 1 to 17 for its “Q4 2026 Business Pulse,” weighting responses by company size.
Marketing services were the most common area reduced or eliminated, cited by 60% of that group. Data analysis or reporting followed at 39%, with bookkeeping or accounting at 38%.
Hiring plans move the other way
The same group also reported stronger near-term hiring plans. Forty-four percent expected to add full-time employees over the next three months, compared with 22% of businesses overall and 8% of businesses with no AI adoption plans. In all, 96% of regular, multi-function AI users planned to maintain or grow full-time headcount.
The broader survey also found rising technology spending. Respondents planning to increase technology budgets, including AI, reached 30% in the fourth quarter, up from 23% in the third quarter. Overall, 83% planned to maintain or increase full-time headcount over the next three months, while 3% planned to reduce it.
Regular, multi-function AI users were more optimistic on revenue as well. Seventy-four percent expected revenue to increase in the next quarter, compared with 54% across the full sample.
What the survey cannot show
Citizens describes the findings as AI changing how small businesses use outside support. The survey itself, however, does not distinguish whether respondents replaced outside providers with AI tools, brought more work in-house with employees using AI or reduced vendor spending for other reasons.
It also does not show what a “reduced” service means in practice. Citizens reported the categories where spending fell or disappeared, but not whether businesses ended contracts, reduced the amount of outsourced work or shifted only specific tasks internally.