Small-business AI adoption is often discussed through surveys about intentions or enthusiasm. JPMorganChase Institute has approached the subject from a different direction: looking at de-identified Chase Business Banking transaction data to identify payments to AI services. That does not tell an owner which product to buy, but it offers a useful view of what businesses are actually paying for and how adoption patterns are changing.
What the JPMorganChase research measures
The Institute's April 2026 report, Understanding the use of AI among small businesses, analyses de-identified transaction data for Chase Business Banking customers from 2019 through 2025. It treats payment for an AI-related service as evidence of adoption and distinguishes that cumulative measure from payment during a particular period.
This methodology has an important limitation that the report itself recognises: free AI use may not appear in payment data. The evidence is therefore best understood as purchasing behaviour, not a complete census of every employee who has opened an AI tool.
New businesses appear more ready to adopt
One of the report's clearest findings is that newer business cohorts adopted paid AI services more quickly than older cohorts did at comparable stages of their lives. The Institute connects the pattern with factors including easier access to generative AI, lower entry costs, cloud delivery and a more mature ecosystem of business applications.
For owners, the useful interpretation is not that a newly formed firm is automatically more sophisticated. New businesses have fewer established systems and habits to unwind. An older company evaluating AI may need to spend more time on integration, process change and staff confidence because the technology is entering an operation that already has ways of working.
Payment patterns suggest movement beyond casual experimentation
The research also reports more consistent payment for AI services and a widening range of service types over time. The Institute interprets those patterns as consistent with businesses moving from sporadic experimentation towards broader operational use, while remaining careful about what transaction data can prove.
This distinction matters. Buying several AI services is not itself a productivity result. A business still needs to ask whether those services remove duplicated work, improve customer handling or support better decisions. The research describes an adoption pattern; it does not guarantee that every adopter receives a return.
Adoption remains uneven between different kinds of firms
The JPMorganChase analysis finds meaningful differences by employer status and industry. Employer firms show higher adoption than non-employers, while knowledge-intensive sectors lead more labour-intensive sectors in the report's data. The wider Small Business in the Age of AI research series also explores differences across owner demographics.
These gaps are a reminder that a generic instruction to embrace AI is not enough. A professional firm working mainly with documents and digital information has different opportunities from a business whose value is created primarily through physical work. The useful application must fit the economics and workflow of the business.
Lower access barriers do not remove implementation work
Affordable, browser-based tools make experimentation easier, but adoption and integration are different stages. Once AI becomes part of routine work, somebody needs to decide which information it can use, how outputs are checked, who owns errors and how the tool interacts with existing systems.
The Institute's implications discuss skills, trusted guidance and effective operational integration as continuing issues. That is a stronger small-business lesson than simply noting that adoption is rising. The difficult part increasingly shifts from gaining access to building a dependable way of working around the capability.
Businesses should interpret research without copying the average
Transaction research can show direction across a large population, but it cannot tell an individual owner that a particular AI purchase is justified. A business should resist buying a tool simply because adoption is accelerating elsewhere. Its own customer journey, information quality, employee capacity and risk exposure remain more important to the decision.
Use external research to challenge assumptions. If comparable firms appear to be adopting faster, investigate what tasks could explain that. If your industry is slower, consider whether the work genuinely offers fewer digital use cases or whether skills and legacy processes are delaying sensible experiments.
Turn the adoption question into an operating question
A useful next step is to choose one recurring task and map it. Identify the inputs, decisions, exceptions and final outcome. Then decide whether AI should assist a person, automate a bounded step or stay outside the process. Run representative cases and measure the complete workload, including review and correction.
This approach keeps the research in proportion. JPMorganChase Institute provides evidence that paid AI adoption among small businesses has accelerated and diversified, while also showing that participation is uneven. The practical opportunity for an individual business lies in what happens after adoption: choosing a worthwhile use case, building the skills and controls around it and integrating the tool deeply enough to improve real work. The power of AI is not unlocked by appearing in an adoption statistic; it is unlocked when the technology becomes a dependable part of an operation that needed improving.