1. Adoption: most owners use AI tools, far fewer run AI in operations
The headline numbers disagree, and the gap is the story. The US Census Bureau's Business Trends and Outlook Survey put AI use at 19.8% of US businesses in May 2026 (U.S. Census Bureau, 2026), with between 20% and 23% expecting to use it within six months. Surveys of small-business owners land much higher: 58% using generative AI in 2025 (U.S. Chamber of Commerce Technology Engagement Center, 2025), 77% of small and midsize businesses using AI regularly (Intuit QuickBooks, 2026) and 75% experimenting with or having implemented AI (Salesforce (Small & Medium Business Trends report), 2025).
The difference comes down to who is asked and what counts as use. The Census figure comes from a large official survey of businesses; the others are surveys of owners and leaders run by companies that sell to them. Read together, they say that trying AI tools is now normal, while putting AI to work in how a business actually operates is still a minority practice.
2. The smallest firms are furthest behind
Size matters. The Census Bureau found AI use below 20% in firms with four or fewer employees, against 37% in firms with 250 or more (U.S. Census Bureau, 2026). Larger firms have staff to evaluate tools, connect them to systems and manage the risk; owner-run businesses usually do not. That is also why the first agents small firms adopt tend to be packaged ones with a monthly plan, such as AI phone receptionists, priced from tens of dollars a month in our 2027 pricing table.
The owners who do use AI report gains: 78% of US SMBs say AI has improved their productivity (Intuit QuickBooks, 2026), and 82% of small businesses using AI increased their workforce over the past year (U.S. Chamber of Commerce Technology Engagement Center, 2025). These are self-reported survey answers, not measured outcomes, but they cut against the idea that small firms adopt AI mainly to cut jobs.
3. What the forecasts say for 2027 to 2029
Gartner expects agents to spread quickly through business software: 33% of enterprise software applications including agentic AI by 2028, and at least 15% of day-to-day work decisions made autonomously by then (Reuters via The Star (reporting Gartner), 2025). In customer service it predicts that agentic AI will resolve 80% of common issues without a person by 2029, with 30% lower operating costs (Gartner (press release republished by CMSWire), 2025).
The same firm also expects a shake-out: more than 40% of agentic AI projects canceled by the end of 2027 (Reuters via The Star (reporting Gartner), 2025). Today's baseline shows how far there is to go: only 1 in 7 customer service issues is fully resolved in self-service (CX Today (reporting Gartner), 2024). Forecasts are not facts; they are the best available view of direction, and they point the same way: agents for routine, well-defined work, and caution about broad projects without clear value.
4. BufferSol view: what this means for a small business in 2027
This section is our opinion as a company that builds AI agents, not a finding from the sources above.
- Start with one workflow you already count. Missed calls, supplier invoices and routine support tickets have clear volumes and costs, so you can prove the result. Size it first with our free calculators.
- Prefer pricing tied to results. Per-resolution and per-minute pricing keeps the cost in line with the work done. Check what counts as a resolution and any minimum commitment.
- Keep a person approving what matters. Refunds, payments and anything a customer will see should have an approval step and a log until the numbers earn trust. This is how a project avoids becoming one of the canceled 40%.
- Buy the packaged version first, build when it stops fitting. A custom agent pays off when the work spans your own systems (ERP, CRM, helpdesk) in ways a packaged tool cannot reach. A 30-day pilot on real data settles the question.