Answer capsule
The SBA Office of Advocacy's September 2025 research spotlight uses Census Business Trends and Outlook Survey data to describe small firms narrowing an AI-use gap. Its accompanying release says the rate for businesses with fewer than 250 employees rose from about 6.3% to 8.8% over roughly six months. That is a population estimate from a defined survey and period, not a target for a particular shop. An owner should choose a first AI job from recurring work, cash and customer consequences, and available review capacity—not from pressure to match an adoption curve.
What the source establishes
- The SBA Office of Advocacy published the research spotlight on September 24, 2025 and analyzes Census Bureau Business Trends and Outlook Survey data.
- The office's release says AI use among firms with fewer than 250 employees rose from about 6.3% to 8.8% over roughly six months and describes small firms as about a year behind large firms on the observed adoption trajectories.
- The report compares uses and investments by firm size and notes that its under-250 size bins were aggregated using survey response counts for weighting.
- The research does not identify whether a particular business should adopt AI, which job it should select, what it should spend, or whether use produced profit, cash, service, or risk improvements for a comparable owner.
Choose a job from the owner's weekly operating record
List recurring work for four weeks: customer questions, estimates, scheduling, order entry, inventory checks, payment follow-up, document preparation, bookkeeping handoff, marketing, and staff coordination. For each job, record frequency, minutes, delay, error and rework, revenue or cash timing, customer consequence, sensitive information, current system of record, and who can verify the result. Shortlist jobs that are repeated, narrow, reversible, supported by clean inputs, and easy for the owner or a trained employee to inspect. Reject a first use that can silently quote the wrong price, commit inventory, move money, file a return, change payroll, disclose protected data, or contact many customers without a dependable approval step. The decision begins with the business's own bottleneck, not an aggregate adoption percentage.
Write a one-page first-job contract
Define the starting trigger, approved inputs, allowed output, prohibited actions, reviewer, response time, customer or employee disclosure where needed, retention, fallback, and stop conditions. Keep the first version draft-only: for example, draft an estimate from an approved price list, summarize a vendor document without changing the books, or prepare appointment options without booking. Require a person to compare the draft with the source and take the external action in the established system. Name the exception route for missing information, unusual terms, complaints, accessibility needs, regulated topics, or a request outside the ordinary service area. A simple contract makes ownership clear and prevents a general-purpose tool from quietly expanding into the whole business.
Price the review burden and the downside
Count subscription or usage charges, setup, data cleanup, integration, staff learning, review, corrections, vendor support, security work, and the time taken from customer delivery. Set an affordable loss limit and identify the worst plausible mistake for the chosen job. Test duplicates, stale prices, missing attachments, wrong customers, ambiguous names, special discounts, returns, late payments, out-of-stock items, altered instructions, and service interruptions. Keep credentials limited and require backups or a manual path. The owner should know how to pause the workflow and reconstruct what happened without calling a consultant. A tool that saves drafting time but adds more checking, customer repair, or bookkeeping cleanup has not yet produced an operating gain.
Compare the pilot with the business, not the benchmark
Run a fixed sample beside the current process and preserve the baseline. Track completed jobs, usable first drafts, review minutes, corrections, missed exceptions, customer delay, conversion or collection timing where relevant, complaints, staff workload, total cost, and any work that shifted elsewhere. Review rejected and low-confidence outputs, not only successes. Continue only if the result improves a declared business measure without weakening service, control, or employee trust. If the job passes, expand one variable at a time—volume, user, customer segment, or action—and recheck. If it fails, close access and retain the lesson. The right outcome may be a smaller use or no current use; neither result is contradicted by a national adoption trend.
Turn this source into a reviewable decision
For AI for Business Owners, use this briefing as a dated decision record rather than a substitute for the source. Preserve AI in Business: Small Firms Closing In, the exact URL, the September 5, 2026 review date, the supported facts above, the editorial interpretation, the limitations, and any buyer-specific evidence. Link that record to the decisions most directly affected: Customer service and appointment support; Quotes, estimates, and proposals; Scheduling and daily operations; Security, privacy, and vendor risk. State whether the source changes the scope, evidence requirement, control, sequence, or only the language used to describe the decision.
Before action, name the accountable owner, affected population and workflow, exact offering or configuration, source data and rights, human decision point, exception and appeal path, complete cost, expected benefit, failure and stop conditions, retained evidence, and next review date. Keep official facts, provider statements, buyer observations, representative tests, measured outcomes, editorial inferences, and unknowns visibly separate. Reopen the record when the source, offer, model, integration, data, policy, population, responsible person, or measured result changes.
Limitations and unknowns
The primary source is a September 2025 SBA Office of Advocacy research spotlight based on Census Business Trends and Outlook Survey data. It supplies population-level estimates and author analysis for specified survey periods; it does not establish current September 2026 adoption, a verified post-cutoff change, causal business outcomes, profitability, tool quality, industry or local fit, or the right operating decision for a specific firm. Survey definitions, response weighting, firm-size grouping, time period, and the distinction between current and planned use matter. The owner's current workload, customers, contracts, finances, systems, data, staff capacity, vendor terms, representative pilot evidence, and qualified accounting, tax, employment, industry, security, privacy, accessibility, insurance, regulatory, and legal advice control.
Decision test
Ask whether the source changes the decision itself, the evidence required, the implementation sequence, or only the language used to describe an existing capability. Record which claims are directly supported, which are provider statements, which require an independent test, and which remain unknown. A source-linked review should make uncertainty easier to see, not bury it inside a blended score.
Questions to take into review
- Which questions have approved answers?
- How does a customer reach a person?
- Which price and scope records are current?
- What changes require owner approval?
- Which constraints and exceptions matter?
- What can change automatically?
- What data leaves the business?
- Who has access and how is it removed?
The publication supports research and executive decision preparation. It does not provide legal, financial, accounting, employment, clinical, cybersecurity, investment, procurement, or implementation advice.