Depending on which survey you read, somewhere between 3.8% and 76% of small businesses “use AI” — and both numbers are accurate, because they’re measuring completely different things. JPMorgan Chase’s transaction-data research puts actual paid adoption at 17.7% as of late 2025. The gap between what gets counted as “using AI” and what actually changes how a business runs is the real starting point for deciding whether — and where — it makes sense for yours.
Every “X% of small businesses use AI” headline this year has cited a real number from a real study. Stack five of them side by side and the range runs from under 4% to over three-quarters. That’s not five studies disagreeing — it’s five different definitions of “using AI” wearing the same headline.
Five Sources, Five Numbers, One Real Explanation
| Source | Reported figure | What it actually counted |
|---|---|---|
| US Census Bureau, Business Trends and Outlook Survey (through most of 2025) | Under 10%, as low as 3.8% | Narrow question: AI used specifically “in producing goods or services” |
| JPMorgan Chase Institute (transaction data, 4.6M businesses, through Dec 2025) | 17.7% | Actual PAID AI subscriptions, verified via banking transaction records — not self-reported |
| Federal Reserve Small Business Credit Survey | 46% (plus 15% planning to start) | Self-reported “uses AI,” small employer firms |
| US Chamber of Commerce | 58% (up from 40% in 2024, 23% in 2023) | Self-reported generative AI tool use specifically |
| Goldman Sachs 10,000 Small Businesses Voices (early 2026, n=1,256) | 76% | Self-reported “use AI in some form” — includes a single one-time trial |
The JPMorgan figure is the one worth anchoring to if you’re trying to gauge where you stand relative to actual competitors, not aspirational ones: it’s the only number here built from what businesses are genuinely paying for, month after month, rather than what an owner remembers trying once. And even that number tells a fast-moving story — firms founded in 2019 took over six years to reach 10% adoption; firms founded in 2025 hit that same mark in six months.
The Loudest Reason for Not Adopting Isn’t Cost
Among small businesses with fewer than five employees who haven’t adopted AI, the SBA Office of Advocacy found that 77–82% give the same answer when asked why: they don’t see how it applies to their business. Cost, security, and technical skill all rank lower.
That’s worth sitting with rather than dismissing. A separate analysis (the AEO Main Street AI Paradox brief) found something that complicates the easy “they’re just wrong” response: small-business owners who report AI as irrelevant to them are, on average, MORE likely to be struggling with basics like taxes, setting up an online presence, or finding customers — not less. The relevance gap tracks real overwhelm, not laziness or technophobia. If you’ve felt like AI coverage is written for businesses bigger or more digitally established than yours, that’s a documented pattern, not just you.
Trust Collapses the Moment the Stakes Rise
A Simply Business survey found 62% of small business owners are already using AI in some capacity — but adoption is wildly uneven by task. Creative and drafting work sees broad comfort. Only 10% of owners said they’d trust AI to handle anything touching their business insurance. Even among active AI users, the same survey found three specific hesitations recurring: accuracy (36%), data security (34%), and not having been shown how to use the tools properly (31%).
That 10% insurance figure lines up with something we’ve covered separately: most commercial general liability policies now let insurers exclude AI-related claims outright, so the instinct to keep AI away from anything insurance-adjacent isn’t overcaution — it’s a reasonable read of where the actual risk sits right now.
A Task-Tier Framework Instead of a Yes/No Answer
“Should my business use AI” is really several smaller questions stacked together, and they don’t all have the same answer. The deciding factor isn’t the task’s importance — it’s whether a wrong output is cheap and fast to catch before it causes damage, the same principle behind the “cheap verifier” idea in our piece on what AI agents can and can’t reliably do.
| Tier | Examples | Why it fits here |
|---|---|---|
| Start here | First-draft marketing copy, social posts, meeting notes, research summaries, routine customer replies | A wrong or weak output is obvious on a quick read and costs you nothing but a rewrite |
| Verify before you send it | Client-facing proposals, pricing quotes, anything published under your name, contracts drafted (not finalized) | Errors aren’t instantly obvious but a human review catches them before real damage happens |
| Hold off / expert review only | Insurance decisions, tax filings, legal documents, HR/compliance calls | Errors are expensive, slow to surface, and the tools and coverage landscape for this category are both still unsettled |
Before You Add a Tool
If the framework above points you toward starting, keep the scope narrow on purpose. Our productivity tools piece covers BCG’s finding that productivity rises with up to three AI tools and drops sharply past that point — pick one task from the “start here” tier, one tool for it, and actually learn it before adding a second. Put a one-page usage boundary in writing before anyone on staff starts using it; our small-business AI usage policy template is built specifically for teams too small for a formal 10-section corporate policy. And before you compare tool subscriptions, price in the hidden costs most sticker prices leave out — we broke those down separately, since the $20/month plan rarely stays $20/month once real usage starts.
Frequently Asked Questions
What percentage of small businesses actually use AI?
It depends entirely on how “use” is defined. Self-reported surveys put it between 46% and 76%. The US Census Bureau’s narrower “used in producing goods or services” measure put it under 10% through most of 2025. JPMorgan Chase’s transaction-data research, tracking actual paid subscriptions rather than self-reports, found 17.7% as of late 2025 — likely the most reliable figure if you want to compare yourself to real competitors rather than aspirational ones.
Is AI actually relevant for a very small business?
Most likely yes for at least some tasks, even though 77–82% of non-adopting micro-businesses say they don’t see how it applies to them. Research also found that owners who feel this way are, on average, dealing with more basic business struggles (taxes, online presence, finding customers) than owners who don’t — suggesting overwhelm rather than an accurate assessment of relevance.
What should a small business avoid using AI for?
Based on where trust is lowest even among active AI users, treat insurance decisions, tax filings, legal documents, and compliance/HR calls as expert-review-only. Only about 10% of small business owners in one survey said they’d trust AI with anything insurance-related, and many commercial liability policies now explicitly exclude AI-related claims.
How many AI tools should a small business use?
Research on this site’s productivity coverage found output rises with up to about three AI tools and falls sharply beyond that. Start with one tool for one task, learn it properly, and only add a second once the first is genuinely part of your workflow.
Why do AI adoption statistics vary so much between reports?
Because each study measures something different: a one-time trial, ongoing self-reported use, use in producing goods and services specifically, or actual recurring paid subscriptions verified by transaction data. None of the numbers are wrong — they’re answering different questions.
Shurah is the founder of AI Tools Daily, tracking pricing, licensing and policy changes across AI tools so readers can make decisions without wading through marketing claims themselves.
