Practical AI training for everyone
LearnModule 1
8 min read
Most owners start with AI the wrong way round. They hear about a tool, try it on whatever is in front of them, get a mediocre result, and decide it is overhyped. The better starting point is your own calendar. AI saves time on a specific kind of work, and until you know how much of that work sits in your week, you are guessing.
For five working days, keep a rough note of what you spend time on, in blocks of fifteen minutes or more. Do not change how you work while you log it. A note on your phone is enough: 'Tue, 40 min, answering delivery questions on WhatsApp'. The point is an honest picture, not a perfect one.
At the end of the week, go through the list and mark every task that has all three of these features. These are the ones worth handing to AI first.
Tasks that fail the third test are not off limits, but they come later and with more care. A refund decision, a contract signature, or a tax figure can all involve AI somewhere in the process. None of them is where you start.
The strongest evidence on this comes from a 2023 experiment with 758 consultants at Boston Consulting Group. On tasks well suited to AI, consultants using GPT-4 finished about 12% more tasks, worked about 25% faster, and produced work rated higher in quality. On a task that looked similar but sat outside what the model does well, the AI users were about 19 percentage points less likely to reach the correct answer than colleagues working without it. They trusted a confident, wrong answer. The researchers called this a 'jagged frontier': the line between what AI does well and badly is uneven and hard to see in advance. Choosing the task well, and checking the output, is the skill. A separate study of customer support staff found AI assistance raised output by about 14% on average, and by far more for newer staff.
One owner's week, sorted
A salon owner logs 6 hours on WhatsApp booking questions, 3 hours on Instagram posts, 2 hours replying to reviews, 90 minutes writing a job advert, and 4 hours on stock orders and supplier calls. The first four are repeat writing tasks with a cheap mistake cost: that is roughly 12 hours of candidate work. Supplier negotiation stays human for now.
Choose the single biggest candidate, not five at once. Note how long it takes today. Use AI on it for two weeks, and time it again, including the minutes you spend checking and correcting the output. That last part matters. A caption that takes two minutes to generate and fifteen to fix has saved you nothing.
Decide before you start what result means keep and what means stop: for example, 'keep it if the task takes at least a third less time and customers notice no difference in quality'. Writing the rule down first stops you from judging the trial by how exciting the tool felt. If the time drops, keep it and move to the next task on the list. If it does not, the usual cause is missing context, such as no voice brief or no standing facts, and the next module fixes exactly that.
It is tempting to multiply the hours saved by your hourly rate and call that the return. It is not, unless the freed time goes somewhere. Ask what actually happens to it: fewer overtime hours paid, a hire you no longer need, more orders handled with the same staff, fewer costly mistakes such as wrong orders or missed invoices, faster replies that win more quotes, or evenings back for you. All of those are real, but only the first four show up directly in the accounts, and 'time I spent scrolling instead' shows up nowhere. Write down where the hours will go, and check a month later whether they did.
Try it now
Start the log today. Write down the last three things you spent more than fifteen minutes on, and mark which of them repeat, are mostly writing or reading, and would be cheap to correct if AI got them wrong.
A few quick questions. You'll see an explanation after each one.
Question 1 of 4
What is the best first step before choosing where to use AI in your business?