The headline figure looks like good news. The SBA Office of Advocacy’s September 2025 research spotlight reports small business AI adoption at 8.8% by August 2025, with the gap to large firms narrowing for the first time in the dataset’s history. Other surveys put the figure higher, the SBE Council 2025 read shows 88% of small businesses using AI tools in some form, and Goldman Sachs’ 10,000 Small Businesses survey found 68% of owners using AI personally. Definitions differ. Direction does not. Small firms are catching up.
The gap that has not closed is confidence. Across surveys of small business owners, only 27% report feeling confident adopting AI, compared with 82% of mid-size firms. That is the real story. Tools are showing up faster than the skills to use them, and the training infrastructure that big employers take for granted does not exist for the corner cafe, the regional accountant, or the four-person agency.
Adoption is the easy part. Confidence is what’s missing
When McKinsey runs the same question across enterprise respondents, the number that stands out is from The State of AI 2025, where 88% of organisations report AI use in at least one function but only 7% have scaled it across the org and only 39% can attribute any EBIT impact to it. Most of those that do report under 5% EBIT contribution. The pattern at the small-business level is the same shape, the licence is bought, the tool is open in a browser tab, and nobody is sure what to do next.
That dynamic shows up in the Anthropic Economic Index too: only about 4% of jobs use AI for 75% or more of their tasks. The “AI does my whole job” narrative is far ahead of reality even in the firms with the most sophisticated rollouts. For a 12-person business with no L&D function, the gap between “we have ChatGPT” and “we use ChatGPT well” tends to widen, not narrow, without deliberate effort.
A training plan that fits a small business budget
Big company training playbooks (six-figure LMS contracts, certified instructor cohorts, dedicated AI champions) do not port down to small firms. Here is a practical four-step plan that does.
Step 1: Pick one task per role. For each person in the business, name one specific task they do every week that AI could help with. A bookkeeper writing client statements. An owner drafting weekly social posts. A salesperson writing follow-up emails. One task each. Resist the temptation to start with a generic course.
Step 2: Pick one tool per task. Most small businesses do not need a tool stack. They need a default. ChatGPT, Claude, and Microsoft Copilot all have free or low-cost tiers that handle 80% of small business writing and analysis tasks. Pick one. Standardise. Reduce the decision fatigue.
Step 3: Build a five-prompt internal library. For each task in step one, write the prompt that produced a usable output and save it where the team can find it. A shared Google Doc works. A Notion page works. The format does not matter. The discipline of “good prompt, save it, share it” is what compounds.
Step 4: Run a 30-minute weekly review. Once a week, the team spends 30 minutes on what worked and what did not. One person shares a prompt that improved an output. One person shares an output where the AI was confidently wrong. Both are valuable. This is where confidence comes from, watching the tool fail in your colleague’s hands as well as succeed in your own.
The BCG and HBS GPT-4 field experiment is the strongest evidence for why step four matters. The 758 consultants in the study completed 12.2% more tasks and 25.1% faster on tasks inside the model’s capability frontier, but performed worse than the control group on out-of-frontier tasks. Knowing where the model fails is as important as knowing where it succeeds. The weekly review is how a small team builds that intuition without paying for a six-figure consulting engagement.
What the data says small businesses should do next
Three findings worth pinning to the wall.
The WEF Future of Jobs Report 2025 puts AI literacy among the top five fastest-growing skill needs through 2030, and projects 39% of workers’ existing skills as transformed or outdated by then. Small businesses competing for talent against larger employers will increasingly need to show new hires that working at the small place will not let their skills atrophy.
The Thryv 2025 small business survey saw AI adoption rise from 39% to 55% year on year, a 41% relative increase in 12 months. The pace of adoption is real, but it is being driven by tool vendors more than by training providers, which means the integration gap (adoption without impact) is the most likely failure mode in the next 12 months.
Goldman Sachs’ 10,000 Small Businesses survey found 68% of owners are using AI personally, but the percentage drops sharply when the question shifts from “do you use it” to “have you trained your staff on it”. The owner is using ChatGPT in their browser. The staff are not.
The fix is not a new platform. It is the four steps above, applied consistently for two quarters. Small businesses are catching up to enterprises on adoption. The ones that catch up on confidence will be the ones that turn the tools into measurable hours saved and revenue earned.