Start with the work that sits waiting for you. A business owner should hand AI the tasks that make other people wait, such as replies, quotes, approvals and summaries, and keep every decision. Audit one working week, pick the single item that holds up the most people, draft it with AI, and read it before it leaves. I call this the Waiting-On-Me Rule.
Most owners do the opposite. They try AI on the work that looks like AI work: a clever marketing idea, a chatbot for the website, a tool somebody recommended at a networking breakfast. Meanwhile three site managers are waiting for a yes, a customer is waiting for an estimate, and the bookkeeper is waiting for an approval that has been in your inbox since Tuesday.
Why do so many owners start with the wrong work?
Because the work that feels like AI work is rarely the work holding the business up.
The smallest firms are where the gap shows. The US Census Bureau's Business Trends and Outlook Survey, a nationally representative survey of US businesses, put AI use at 19.8% of firms as of 3 May 2026. It reported that use had risen among firms with at least 20 employees, while it had not changed significantly among smaller ones, and that fewer than 20% of firms with four or fewer employees used AI in any business function (Census Bureau, published 26 May 2026, data collected 14 December 2025 to 3 May 2026). In a firm of four, the owner is the management layer. Nobody else can run the experiment.
The owner's time is also the scarce thing. In a survey of 1,000 UK micro, small and medium-sized business owners by American Express and Small Business Saturday UK, reported on 20 July 2026, owners estimated they spend an average of 11 hours a week on administrative or finance tasks, about six working days a month, against 3.6 days a month on sales and business development. The Federal Reserve Bank of San Francisco, summarizing the 2024 Small Business Credit Survey in a note published 1 September 2026, lists time and capacity among the barriers to AI use for small US firms, alongside regulation and knowledge gaps.
The British Chambers of Commerce adds a sobering detail from its January and February 2026 survey of 668 UK firms, 94% of them with fewer than 250 staff: 54% of firms were using AI, and 95% of the SMEs using it reported no change in workforce size over the past year (BCC and the University of Essex, published 19 March 2026). Nobody is reorganizing around AI yet. The gain, where there is one, arrives as hours, and in a small firm those hours first belong to the owner.
Put those together and you can see the trap. The owner has no time to learn AI because the owner's time is eaten by the work AI could draft. That is my reading, not the surveys' finding, but it explains why advice to "experiment with AI" so often dies in the second week.
What is the Waiting-On-Me Rule?
The Waiting-On-Me Rule is where I suggest an owner starts with AI: not with the work the owner enjoys, and not with the work that looks most impressive, but with the work other people are waiting on. The owner drafts it with AI, decides it personally, and measures success by how many people stop waiting.
Three reasons make waiting the right place to start.
First, waiting multiplies. When you take an hour on a reply, the hour is not the only cost. Everyone downstream of that reply loses their own hour too, and in a small firm they have nobody else to ask.
Second, waiting work is usually writing, sorting or summarizing: the replies, the estimate letters, the notes from a call, the Sunday-night list of what is stuck. AI is good at a first draft of exactly those. It is much less good at deciding which client to keep or what to promise, and you do not want it deciding those anyway.
Third, your team sees it. When the owner's backlog shrinks, everyone feels the difference within days, and nobody has to be convinced by a slide deck. My governing principle is to outsource the doing, not the thinking. The Waiting-On-Me Rule applies it to the one person who cannot hand the thinking to anyone else.
How do you audit your own week?
You keep a log for five working days, ten minutes at a time, and score each row by who is waiting. Here is the sheet. Copy it into a notebook or a spreadsheet.
| Task | Who is waiting | How long they wait | Could AI draft it? | What I still decide | Check by |
|---|---|---|---|---|---|
| Reply to a supplier query | Two site managers | 1 to 2 days | Yes | Whether to accept the terms | Friday |
Fill a row whenever you notice you are the reason something has stopped. Do not wait until Friday to remember. By then the Tuesday item has vanished from memory and only the Thursday annoyance remains.
On Friday, give each row a waiting score: the number of people waiting multiplied by the days they waited. The row with the highest score is your starting point, even if it is dull. Then sort the rest with this table.
| If the task is... | Then... |
|---|---|
| Waiting on you, and mostly writing, sorting or summarizing | Start here. AI drafts, you decide. |
| Waiting on you, and mostly judgment (whom to hire, which client to keep, what to promise) | Keep it. Use AI only to prepare the facts. |
| Not waiting on anyone, but you dislike it | Later. Dislike is not a priority. |
| Contains client or staff personal data | Check your AI policy before anything is pasted. My UK small-business AI policy template shows what to cover. |
What does it look like in a real week?
Take a hypothetical 14-person electrical contracting firm. The numbers below are invented for illustration and are not data.
The owner logs her week. Four customers are waiting three days each for an estimate letter, a score of 12. Two site managers wait a day for answers to supplier queries, a score of 2. The bookkeeper waits two days for approvals, a score of 2. Her own Sunday-night summary of what is stuck, which nobody waits for, scores zero and takes her an hour.
The estimates win. She dictates her site notes after each visit, asks AI to turn them into a draft estimate letter in the firm's usual structure, and then checks the scope, the exclusions and the promise before it goes out. She still decides what the firm will commit to. The AI decides nothing.
After four weeks she counts one thing: how many days a customer waits for an estimate. If the number has not moved, the rule has told her something useful, and she changes the draft routine or the review habit, not the tool.
What should stay with you?
Anything where being wrong is expensive and hard to undo: whom you hire, which customer you let go, what the firm promises, what it signs. AI can prepare the facts for these. It should not hold the pen.
The second half of the answer is how you review a draft, because the owner reading AI output at ten at night is the weakest safeguard in the building. I use the Loop Test for this. The Loop Test is three questions I ask before accepting "a human checks it" as an answer: Would that person notice a wrong result without being told to look for it? Would they notice in time to stop it mattering? And if they missed it, could the work be put back? Oversight that survives all three is a safeguard. Oversight that fails any of them is a signature.
Apply it to yourself. An estimate letter you read at your desk with the site notes beside it passes. One you skim on your phone in the car park does not. Choose the review that passes.
What I Tell Business Owners
Across the business teams I train, owners ask me which tool their people should use far more often than they ask what they should hand over themselves. That is a composite pattern, not a count. It matters because the owner sets the pace.
The one business story I can tell in detail is on my business page. 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 business what was possible. After the September 2024 session the IT Director wrote: "It was a great session and I've heard a lot of good feedback from the attendees." I do not take from this that AI spreads on its own. I take from it that someone at the top asked for more.
How will you know it is working?
Count one number and ask one question. The number is days waited on the one task you chose, before and after. The question goes to the people who were waiting: has anything changed for you in the last month?
If both answers are good, take the next row on your sheet, not the next tool. If you are not sure the habit has stuck across the team, the Reach, Rhythm and Role Check in Is Your Team Really Using AI? A Depth Check and Sheet gives you three counts to take. And once the owner has a working routine, the next question is what the team needs, which What AI Training Does Your Team Need? A One-Page Brief answers. For the finance and admin rows on your sheet, see my guide to AI for finance and admin in a small business.
What should you do this week?
- Monday morning, copy the sheet into whatever you already use and keep it open.
- For five days, add a row each time you realize you are the reason something has stopped.
- On Friday, score the rows and choose the one with the highest waiting score.
- Before you use AI on it, check what your policy allows you to paste, and what it does not.
- For two weeks, draft that one task with AI and run the Loop Test on every draft before it leaves.
- Put a date four weeks out in your diary to count the days waited and ask the people who were waiting.
- Tell your team what you did and what changed. That is the cheapest training you will give them this year.
Where to go next
If you want a leadership team to agree where AI starts before anyone spends another month experimenting, take a look at my business strategy and governance work. If you would like to see where your business stands first, the Business AI Readiness Scorecard takes about five minutes.
Sources and further reading
- U.S. Census Bureau, "Large Firms With at Least 20 Employees Biggest AI Users", Business Trends and Outlook Survey, 26 May 2026: census.gov
- Federal Reserve Bank of San Francisco, "AI Adoption Among Small Businesses: Qualitative Insights from the Small Business Credit Survey", 1 September 2026: frbsf.org
- American Express and Small Business Saturday UK, SME Business Barometer, as reported by Retail Times, 20 July 2026: retailtimes.co.uk
- British Chambers of Commerce and University of Essex ISER, Future of Work: AI in the Workplace Report, 19 March 2026: iser.essex.ac.uk
Dan Fitzpatrick, The AI Educator, is a Forbes contributor, international keynote speaker and bestselling author who has trained more than 150,000 people across 30+ countries. About Dan
