Interest in artificial intelligence is changing how small firms evaluate administrative tasks, client communication and software spending. Research from the JPMorganChase Institute provides an evidence-based way to discuss adoption, but its scope must be respected: transactions indicating paid AI service use do not prove that the business achieved a particular profit, productivity or customer-service outcome.
Read the underlying research rather than its headline
JPMorganChase Institute's study Understanding the use of AI among small businesses examines AI-related spending in a defined US Chase Business Banking sample, covering observations from 2019 through 2025. It is not a census of every small business in the US or a direct measure of UK firms.
Understand the study's adoption definition
Payment to an AI provider can indicate that a firm adopted a paid service under the study methodology. It may not capture free-tool experimentation or reveal how intensively a subscribed product was used. A business paying for an AI tool is not automatically a business receiving measurable value from it.
Distinguish adoption from business performance
AI may support drafting, sorting or reviewing information, but the transaction dataset alone cannot show that a business reduced downtime by a named percentage or achieved a stated financial return. Avoid attributing invented bakery case studies, savings or endorsements to JPMorganChase researchers.
Consider business differences
Employer firms, non-employer firms and newer enterprises can display different patterns of paid service adoption. Industry and revenue mix may also influence technology purchases. Those differences describe associations in the sample and do not establish a universal implementation rule.
Start with one bounded operational task
Choose a recurring activity that staff can verify, such as preparing first-draft routine replies from approved text or summarising meeting actions. Describe the present process and record baseline errors and time spent. Avoid beginning with sensitive decisions or granting a service more information than needed.
Check accuracy and exception handling
Test both typical and difficult examples. Record unsupported claims, incorrect outputs and cases requiring escalation. Assign a responsible reviewer, especially for financial, legal, contractual or client-specific communications. If error correction outweighs time saved, the pilot has not established a benefit.
Review privacy and contractual controls
Consider which business and personal information will be sent to the service, the vendor's current terms, retention and access controls, and any contractual confidentiality obligations. Use a controlled trial dataset where possible and document approved use by staff.
Measure net effect rather than headline speed
Compare completion time, review effort, rework, consistency and customer effects with the existing workflow. Include subscriptions, setup, integration and training costs. An impressive generated draft is not enough evidence for rollout without dependable outcomes.
Plan continued oversight
Model behaviour and available features can change. Re-evaluate representative tasks after updates, provide staff a way to report problems and keep a human owner for important commitments. A tool should support accountable employees rather than conceal responsibility.
Use the report for context, not endorsement
The JPMorganChase Institute research offers useful information about defined paid-AI adoption patterns. It does not endorse a particular vendor, verify a fictitious local business result or establish that AI will automatically transform every small firm's operations. Base investment decisions on your own measured pilot and verifiable current sources.
Frequently Asked Questions
Does JPMorganChase Institute research measure every US or UK small business?
No. The cited study uses a defined sample of US Chase Business Banking transactions. It cannot be treated as a universal UK adoption rate or a census of all US firms.
Does paying for an AI service prove a business became more productive?
No. Payment-based adoption measures indicate purchases under the study definition. Businesses need separate workflow evidence to evaluate accuracy, staff effort, costs and real outcomes.
How should a small firm start assessing AI?
Choose one bounded task, protect sensitive information, trial representative cases with human review and compare total cost and corrected results against the existing process before expanding.