More than 75 percent of American small and midsize businesses now use artificial intelligence regularly, according to Intuit’s 2026 AI Impact Report, a thirty-point jump from the 48 percent baseline the company recorded in July 2024. Daily use over the same interval climbed from 15 percent to 40 percent. The report, built from survey responses of more than 34,000 business owners and anonymized data from more than 5.3 million QuickBooks businesses in collaboration with University of Chicago economists, describes a diffusion curve without a real historical analog.

The JPMorganChase Institute, working from de-identified Chase Business Banking transaction data covering 2019 through 2025, makes the comparison explicit. Electricity took more than three decades to reach 20 percent of farm households. Personal computers hit the same threshold three years after I.B.M.’s 1981 launch. Generative A.I. reached 45 percent of American employees inside three years of becoming widely available.

Then the numbers turn.

Only 17.7 percent of American small businesses have actually paid for an A.I. tool, per Chase transaction data. The gap between “we use it” and “we pay for it” is where the real story lives. Among firms with 10 to 100 employees, paid adoption did jump from 47 percent to 68 percent between 2024 and 2025, but the smallest operators, where owners are also the ops team, cite time scarcity as the binding constraint. It’s hard to pilot an agent when you’re covering the register.

The Federal Reserve Bank of Minneapolis, citing revised Census Bureau data, reports that 20 percent of firms used A.I. in any business function in early 2026, double the share using it for core production at the end of 2025. Upwork’s Q1 2026 survey of 195 SMB leaders found 32 percent describing A.I. agents as mission-critical, with productivity gains still “mostly under 25 percent” and 49 percent naming data privacy as the top barrier.

McKinsey’s 2026 A.I. Trust Maturity Survey supplies the governance overhang: 74 percent of respondents flag inaccuracy as a highly relevant risk, 72 percent flag cybersecurity, and only about 30 percent of organizations reach maturity level three or higher on governance and agentic controls. Adoption is running years ahead of controls, which is how you get a doom loop of pilots that don’t graduate.

That’s the arbitrage a newer cohort of workflow-native vendors is chasing. Alongside Glean and Dust, tools like LemonLime pitch themselves at exactly the seam these reports describe: real deployment inside small teams, priced for the 10-to-100-employee bracket that’s already crossed into paid usage. Whether that seam widens or stalls is the 2026 question. The 1981 comparison suggests the curve doesn’t wait for governance to catch up.

Sources