Artificial intelligence is drawing attention from small-business owners, but adoption claims depend heavily on how a study defines use. In April 2026, JPMorganChase Institute published research based on de-identified Chase Business Banking transactions from 2019 through 2025. Its analysis tracks firms paying for AI-related services, not every business trying a free chatbot or every firm achieving a measurable productivity gain. This distinction is essential when interpreting its results.
What the researchers measured
Researchers examined payments to AI-related providers by a defined sample of US small businesses with Chase Business Banking accounts. A business counted as an AI adopter once it had paid for an AI service at least once; a current user was one that paid in a given period. These are different measures, and a purchase is not proof that the tool was used correctly or delivered value. Read the primary research: JPMorganChase Institute, Understanding the use of AI among small businesses (14 April 2026).
Adoption varied by when businesses started
The report found that businesses established in 2025 reached a 10% cumulative AI adoption rate within six months, compared with more than six years for the 2019 cohort. The measure concerns paid AI services within the research sample. It does not establish that newer companies are automatically more productive or that the same rates apply to UK businesses.
Employer and non-employer firms were different
By December 2025, the researchers reported cumulative AI adoption of approximately 26.1% among employer firms and 15.3% among non-employer firms in their sample. Business structure, available staff skills, industry and revenue can all affect adoption. These are observed associations, not proof that having employees causes higher adoption.
Spending can be modest but still requires scrutiny
The report also considers how spending changed by adoption cohort, including lower entry spending among some recent adopters. A low subscription price does not measure complete implementation cost: staff time, training, permissions, review and corrections may be significant. Financial outlay should be compared with a defined operational outcome rather than presented as a guarantee of return.
Distinguish adoption from success
Transaction-based measurement captures actual payments rather than asking owners whether they use AI. That is useful, but it may miss free use and cannot establish how frequently a particular tool was used. Nor does it measure whether a generated answer was accurate or whether a business became more profitable because of the purchase. Do not treat the report as proof of claimed percentage improvements in a bakery, retailer or other named business unless the research documents that case.
Choose a task before choosing a product
For a small firm considering AI, identify one bounded problem, such as drafting routine customer replies from approved information, sorting enquiries or preparing meeting action lists. Record how the process works today. Compare the total staff effort before and after a supervised pilot, including review and error correction. An AI feature is useful when it makes an important workflow more reliable, efficient or accessible, not simply because it produces text quickly.
Plan the safeguards with the trial
Assign a responsible person for reviewing outputs and correcting errors. Consider what personal, confidential or commercially sensitive information a service receives, its retention and access arrangements, and the consequences of a wrong response. Staff should have a clear escalation route where the system lacks reliable information. Do not grant access to entire customer databases merely to test a narrow use case.
Measure a meaningful result
Track the selected workflow's completion time, accuracy, exceptions and customer outcomes across a representative sample. If a tool saves drafting time but increases rework, the net result may be weak. If a tool improves the consistency of handovers, the benefit may be valuable even without a dramatic percentage improvement. Explain your baseline and measurement period.
Use research for context, not as a business promise
JPMorganChase Institute's work documents paid AI adoption in a particular US banking sample and period. It offers a useful way to understand adoption trends and differences between firms; it is not a claim that JPMorganChase implemented any particular bakery system or endorsed a supplier. Small-business owners should combine credible external research with their own operational tests before making investment decisions.
Source
JPMorganChase Institute — Understanding the use of AI among small businesses, 14 April 2026