A business can give every member of staff an AI login on Monday and be exactly the same business on Friday. Most small businesses are not getting value from AI yet because they have stopped at access. Value arrives only when AI is attached to a specific job, done a shared way, inside a routine that has actually changed, with someone checking whether the result is better. In a March 2026 Goldman Sachs survey of small businesses in its 10,000 Small Businesses program, 76% said they used AI. Only 14% said it was fully embedded in their core operations.
That gap is not a failure of the software. It comes from six habits, and each one feels like progress while you are doing it.
Is the problem the technology or the way it is used?
It is mostly the way it is used: the businesses getting value from AI differ from the rest in how they organize work, not in which tools they bought.
Start with small and medium-sized businesses (SMBs, or SMEs in the UK and Europe). The Goldman Sachs 10,000 Small Businesses survey (March 17, 2026; 1,256 participants) found that 73% "would benefit from additional access to training and implementation resources." The US Census Bureau (May 26, 2026) reported that between December 2025 and May 2026, AI use "increased among firms with at least 20 employees but didn't change significantly among firms with fewer than 20 employees." In the UK, the British Chambers of Commerce (March 18, 2026) reported that "more than half of UK firms (54%) are now actively using AI," up from 35% in 2025, yet a University of Essex working paper on the same survey found that only "around one in ten firms have adopted bespoke AI implementations."
Larger companies show the same shape. BCG's "The Widening AI Value Gap" (September 2025, more than 1,250 firms worldwide) found that only 5% are "achieving AI value at scale" and that "fully 60% of companies are not achieving material value at all." In McKinsey's "The state of AI in 2026" (August 25, 2026), "nearly nine in ten respondents report regular use of AI in at least one business function," yet only 37% attribute at least some EBIT impact (roughly, operating profit) to it.
My reading: use is common and value is rare. The two studies that ask what separates the few from the many, BCG's and McKinsey's, find the difference in workflows, skills and leadership, not in the software. That is good news for a 20-person firm. You cannot outspend a bank on technology, but you can change a routine by Friday.
The Freewheel Diagnostic: where does your AI stop turning the business?
The Freewheel Diagnostic finds the exact point where AI stops connecting to real work, by asking four questions about one job at a time.
A freewheel is a gear that spins without driving anything. AI access on its own looks just like that: busy, fast and connected to nothing. Here is the definition I work from:
The Freewheel Diagnostic is four questions I suggest a business asks about any job where AI is meant to be helping. Is it a named job with an owner? Is there a shared way of doing it with AI? Has the routine around it changed? Does someone check whether the result is better? A job that fails any question is freewheeling: the AI is spinning, but nothing in the business turns. Value arrives through the jobs that pass all four.
Take a hypothetical 25-person property management firm where everyone has Microsoft Copilot or ChatGPT. Run the diagnostic on writing rental listings. A named job? Yes: the leasing team writes them and the office manager owns quality. A shared way? No. Six people prompt six different ways, and two do not use AI at all. The firm stops there, because the first "no" is where the gear slips.
That first "no" also tells you which of the six mistakes below you are making.
Mistake 1: Treating access as adoption
Giving everyone an AI tool is the start of adoption, not the end of it.
Why it happens. Access is the one step an owner can finish alone, in an afternoon, and point to afterward. It looks like a decision.
What it costs. Use stays private and uneven. A few people get faster; the business does not, because nothing it does has changed. The distance between Goldman Sachs's 76% and 14% is this mistake, measured.
Instead. Treat access as the starting line. The next move is not a second tool. It is naming the jobs.
Mistake 2: Letting everyone find their own uses
When nobody names the jobs AI is for, the gains stay in the heads of two or three enthusiasts.
Why it happens. The tools can do almost anything, so owners hesitate to narrow them. Staff, meanwhile, are often unsure what they may put into a chatbot, so the cautious ones hold back.
What it costs. The best method in the building lives in one person's browser history. It never spreads, and it leaves when they do.
Instead. Name three jobs where AI should help first, and say plainly what is and is not allowed. A one-page statement is enough to start; whether a small business needs an AI policy covers what it should say, and how a small business should start using AI covers choosing the jobs.
Mistake 3: Training people on the tool instead of their work
A demo of features teaches people what AI can do; only practice on their own work teaches them what it should do here.
Why it happens. Tool training is easy to book. A lunchtime tour of what ChatGPT can do feels productive, and everyone leaves impressed.
What it costs. Impressed is not the same as changed. On Monday people face their real inbox, their real quotes and their real customers, and nothing in the demo looked like those. When 73% of Goldman Sachs's respondents ask for more training and implementation support, I read that as training that has not yet reached the work.
Instead. Train on the named jobs, with realistic examples, until each person has done the job with AI at least once in the room. What makes AI training for small business teams actually work goes into how.
Mistake 4: Bolting AI onto the old routine
If the routine around a job stays the same, the time AI saves leaks away before it reaches the business.
Why it happens. Adding a tool to a step is easy. Changing who does what, in which order, and who signs it off means an awkward conversation.
What it costs. Picture a proposal whose first draft now takes ten minutes instead of an hour, then waits three days in the same approval queue. Nobody lies about the time saved; it simply evaporates. McKinsey found that "nearly three-quarters of high performers report fundamentally redesigning workflows because of their AI use," compared with "just one-quarter of other respondents." BCG's rule of thumb agrees: "70% of a business's strategic focus should be on the people and processes, 20% on the tech, and 10% on algorithms."
Instead. Redraw the job on paper. Which step disappears? What does the human now check, and when? Who can approve it sooner?
Mistake 5: Running pilots that never have to end
A pilot without a decision date becomes a permanent experiment that nobody owns.
Why it happens. Pilots feel safe. They promise learning without commitment, and nobody has to say no to anything.
What it costs. Energy drains away. The enthusiasts move on, the skeptics conclude AI was a fad, and the next idea arrives to a tired team.
Instead. Before anything starts, ask the Exit Question: "if this works, what will change, who has already agreed to it, and on what date will you decide? A pilot without an answer is not a pilot. It is an experiment with no exit."
Mistake 6: Counting logins instead of results
Usage tells you people opened the tool; only a before-and-after measure on the job tells you whether the business gained.
Why it happens. Usage is the number the software reports for free. Results take a stopwatch.
What it costs. You cannot tell which jobs to scale and which to drop, so everything carries on at half speed. McKinsey found its high performers twice as likely as others "to report that their organizations have defined processes to measure the impact of those initiatives."
Instead. Give each named job one measure you already track: turnaround time on quotes, errors caught in invoices, hours to answer a customer. Record it before, then four weeks after. Evidence is one of the seven dimensions in how ready your business is for AI.
What I See When I Train Business Teams
The gap between access and value usually shows up first as a gap between the people at the top and the people doing the work.
Across the business teams I train, leaders tend to describe AI as something they have rolled out, while staff describe it as something they are quietly trying on their own. Both are telling the truth. The two accounts converge fastest when the owner uses AI on their own work in front of the team, including the drafts that come out wrong. McKinsey's high performers are "twice as likely as others to say that their senior leaders demonstrate commitment to AI initiatives," and BCG reports deeply engaged C-suites at nearly all future-built companies against 8% of laggards.
More than 500 keynotes in 30 countries across five continents have taught me that the room changes and the pattern does not. The questions that reach my daily podcast and my newsletter, read by more than 44,000 subscribers, rarely ask which tool to buy. They ask some version of this: we have it, so why does nothing feel different? The Freewheel Diagnostic is my answer.
What should you do this week?
Pick one job, run it through the four questions, and fix the first "no" before you add anything new.
- Ask your team where they already use AI. A fifteen-minute conversation usually surfaces three or four real jobs.
- Run the Freewheel Diagnostic on each job. Mark the first "no."
- Choose the job closest to passing. Write its shared method on one page: what goes in, how to prompt, what a person checks before it leaves the building.
- Change one step in the routine. Remove a handoff, move a check earlier or shorten an approval.
- Pick one measure and a review date. Put the date in the calendar now.
Then hold the plan to the Monday Test: "could a member of staff read it and know what to do differently on Monday morning?" If yes, one job has moved from access to value. Next month, do the next one.
Where to go from here
If your team has access to AI but nobody has shown them how to use it well on their real work, that is exactly what my AI training for teams is built for. If the gap starts higher up, with no agreed answer to what AI is for in your business, begin with a leadership session on AI strategy and governance, or take the free Workplace AI Readiness Check to see where you stand.
Sources and further reading
- Survey: Small Businesses Embrace AI, But Need Training and Support to Fully Harness It, Goldman Sachs 10,000 Small Businesses Voices, March 17, 2026.
- Large Firms With at Least 20 Employees Biggest AI Users, US Census Bureau (Business Trends and Outlook Survey), May 26, 2026.
- Half of SMEs Using AI, With Limited Headcount Impact So Far, British Chambers of Commerce, March 18, 2026.
- AI adoption and workforce change in SMEs, David Bharier, Ben Etheridge and Paulo Morais, ISER Working Paper 2026-01, University of Essex, March 18, 2026.
- The Widening AI Value Gap: Build for the Future 2025, Boston Consulting Group, September 2025.
- The state of AI in 2026: On the road to ROI, McKinsey & Company, August 25, 2026.
Dan Fitzpatrick is The AI Educator: Forbes contributor, international keynote speaker and bestselling author who has trained more than 150,000 people across 30+ countries. More about Dan.
