Depending on whose survey you read, either three in four US small businesses use AI or fewer than one in five do. Both numbers are honest, and the gap between them is exactly what a small business AI plan for next year is for. That plan should fit on one page and hold only lines that name a job, a person and a date: the jobs where AI already works, the one or two jobs it moves into next, and the things AI will never do in your business. Anything else is a wish.
October is the month to write it. Budgets for next year are being set now, and once a budget sets, whatever it says about AI (often a software line and nothing more) is what the business will do for the next twelve months.
Why do the AI adoption numbers disagree so much?
They disagree because they count different things: some count people using AI at all, others count businesses using it in their work, and in small firms those are not the same.
Intuit's 2026 AI Impact Report, published on 12 May 2026 and drawing on more than 34,000 owner responses, found that "77% of US businesses now use AI regularly, up from 48% in July 2024." Intuit counts "any kind of AI use, from free tools and built-in features to dedicated paid subscriptions."
The US Census Bureau tells a different story. Its Business Trends and Outlook Survey asks firms whether they used AI in any business function in the past two weeks. In a report published on 26 May 2026, Census Bureau authors Adam Grundy, Cory Breaux and Dhanapati Khatiwoda found overall AI use hovering between 17% and 20% from December 2025 to May 2026. Use rose among firms with at least 20 employees over those six months but "didn't change significantly among firms with fewer than 20 employees." Fewer than 20% of firms with four or fewer employees reported using AI at all.
A Federal Reserve note by Jeffrey Allen, published on 3 April 2026, set the Census figure of 18% beside 78% from a survey of executives, weighted by employment, and 41% from a survey of individual workers. Much of the gap, Allen explains, comes from who gets counted: large firms adopt far more, and surveys that lean toward large employers or ask individuals report far higher numbers.
The UK looks similar from another angle. The British Chambers of Commerce, working with Atos and the University of Essex, reported on 18 March 2026 that 54% of UK firms are now actively using AI, up from 35% a year earlier. Among the firms using it, 86% said job roles had remained unchanged.
Here is my reading, kept apart from what the sources found. In most small businesses, AI has arrived through people, not through the business. The owner drafts emails with it. Someone in marketing writes posts with it. The office manager asks it to summarize a long supplier contract. Ask those people and they use AI. Ask whether AI is part of how the business produces its quotes, its job sheets or its month-end, and the honest answer is often no. The roles stay the same because nobody has redesigned the work. A plan for next year is how a business moves AI out of personal habit and into the way the work runs.
What goes wrong with the AI line in most business plans?
Most AI lines in a business plan name a tool, a percentage or a hope, and none of those tells anyone what to do differently.
When I work with leadership teams on next year's plan, the AI line in the first draft is almost always one of three kinds:
- The tool line. "Roll out Copilot to all staff." It names something to buy, not something to change. Six months later everyone has a license and the work looks the same as it did.
- The percentage line. "Use AI to cut admin time by 20%." It sounds measurable, but there is no job attached, no starting figure and nobody whose time it is.
- The hope line. "Become an AI-first business." It is a mood. Nobody can fail it, so nobody can succeed at it either.
All three fail a question I put to any AI strategy. The Monday Test is a single question to ask of any AI strategy before it is signed off: could a member of staff read it and know what to do differently on Monday morning? Read "roll out Copilot to all staff" to your office manager and ask what she will do on Monday. She will log in. Then what?
The Wish Test
The Wish Test is the check every line of a small business AI plan should pass before it goes in: it must name a job, a person and a date.
The Wish Test is how I suggest a small business checks its AI plan for the year ahead: every line must name a job, a person and a date. A line that names only a tool, a percentage or a hope is a wish, not a plan. Sort the lines into three kinds, proven, next and never, and if a line cannot name its job, its person and its date, rewrite it until it can or strike it out.
It is a way of thinking I am proposing, not a research finding. The three kinds of line do different work.
Proven lines hold the jobs where AI already works in your business, and say who keeps them working. "Already works" means someone measured the job before AI went in, and it got faster without coming back for more corrections. The method is in the Before Number Rule. Plenty of businesses will discover they have no proven lines yet. Better to learn that in October than next September.
Next lines name the jobs AI moves into during the year, one at a time, each with an owner, a start date and a decision date. This is the One Job Rule repeated across a year: choose one job, not one tool, prove it, then choose the next.
Never lines say what AI will not do in your business next year, and who holds that line. They are the shortest lines on the page and often the first ones staff read, because they answer the question people are too polite to ask out loud: what is safe? They sit beside your AI policy, not in place of it.
The date matters as much as the job. A job without a date drifts into next year's plan unchanged. A date without a decision is a calendar reminder. So every next line carries two dates: when AI starts in the job, and when you will decide whether it stays.
What Good Looks Like
A good small business AI plan is one page and six to ten lines, and every line could be read aloud at a team meeting without anyone asking "meaning what?"
Take a hypothetical 25-person heating and plumbing firm planning its year in October. Its first draft said "invest in AI tools and train staff." After the Wish Test, the page reads like this.
Proven
- Quote follow-up emails are drafted with AI and checked by whoever did the survey visit. Owner: the office manager, who keeps the time and correction figures each month and reports them at the January meeting.
Next
- January to March: engineers' voice notes become typed job sheets with AI. Owner: the service manager. Starting figures taken by 15 January; decision on 31 March.
- April to June: first drafts of tender responses for housing association contracts. Owner: the managing director. Starts 1 April; decision on 30 June.
- July to September: chosen at the June review from a shortlist the team writes. Owner: the office manager, who gathers the shortlist by 15 June.
Never
- AI never sends a customer anything about safety or money without a person reading it first. Owner: the managing director, reviewed at every quarterly meeting.
- No customer's personal details go into a tool that is not on the approved list. Owner: the office manager, who reviews the list on 1 March.
Notice what is missing. There is no software line on the page, because buying follows the jobs, not the other way round. There is no percentage target, because the starting figures will produce real ones. And the third quarter is left open on purpose, so that what the firm learns in the first six months gets to choose it.
One business I have worked with shows what that move looks like. In April 2023 I spoke about AI in marketing at a business event in Penrith. A year later, the marketing team at theidol.com was using AI for content writing, and the managing director asked me back to show teams across the whole business what was possible. After that session in September 2024, the IT Director wrote: "It was a great session and I've heard a lot of good feedback from the attendees." That second invitation is the moment a plan exists for: one team has proved something, and the owner wants it to become how the whole business works.
How fast should a small business move AI into new jobs?
For most businesses under 50 people, one new job a quarter is the right pace, and faster only once the jobs you have already proved run without you.
The case for speed is real. Competitors are moving, tools are cheap to try, and four jobs in a year can feel timid when every member of staff already uses AI for something. If you already have proven lines, a named person keeping them, and a team asking for more, take two jobs a quarter.
The case for patience is that every job needs the same three things before it counts: a brief clear enough that a new hire could follow it, a person who checks the output, and a starting figure to compare against. Small teams run out of attention long before they run out of ideas. A plan with twelve next lines usually ends the year with none proven, which leaves the business worse off than a plan with four.
Who keeps the plan matters as much as its pace. In most small firms the owner should hold the never lines, and an AI steward should hold the rest: someone the team already asks for help, with protected time to do it.
What should you do this week?
You can draft next year's AI plan in an afternoon, without a consultant or a data team.
- Ask everyone one question, in person or in a two-line message: "What has AI saved you this month?" Write down every answer, including "nothing."
- Sort the answers into three piles: uses with a job and some evidence behind them (possible proven lines), uses worth turning into a proper job (possible next lines), and uses that make you uneasy (possible never lines).
- Draft one line for each in plain words, then run the Wish Test on every line. Rewrite or strike out anything that cannot name its job, its person and its date.
- Choose the first next line for January and name its owner now, so the starting figures are taken before the holiday rush blurs them.
- Read the page aloud to someone who was not in the room and ask what they would do differently on Monday. If they cannot say, rewrite.
- Take the page into the budget meeting before the software budget is agreed, not after.
Where to go next
If your leadership team needs to agree what AI is for in the business before anyone writes a plan, that is the alignment work I do with leadership teams through AI strategy and governance for business. If you would rather see where you stand first, the free Business AI Readiness Scorecard takes about five minutes and ends with a 90-day plan. And if you are planning a leadership away day or a members' event for the new year, I speak on exactly this in my keynotes for business audiences.
Sources and further reading
- Intuit, "2026 AI Impact Report: Mapping Adoption, Use, and Impact Across Small to Midsize Businesses", 12 May 2026. intuit.com
- Adam Grundy, Cory Breaux and Dhanapati Khatiwoda, "Large Firms With at Least 20 Employees Biggest AI Users", US Census Bureau, 26 May 2026. census.gov
- Jeffrey S. Allen, "Monitoring AI Adoption in the US Economy", FEDS Notes, Board of Governors of the Federal Reserve System, 3 April 2026. federalreserve.gov
- British Chambers of Commerce, "Half of SMEs Using AI, With Limited Headcount Impact So Far", 18 March 2026. britishchambers.org.uk
Dan Fitzpatrick is The AI Educator: a Forbes contributor, international keynote speaker and bestselling author who works with business leadership teams on AI strategy, training and governance. More about Dan.
