Name one job your business used to do that it no longer does, or one it now finishes far faster, and say who checks the result. If you can name that job, you are using these tools strategically. If you cannot, you are still practicing.
You have had a paid account for a year, you open it most days, and some answers have impressed you. And if a friend asked you over lunch what any of it has done for the business, you still could not say.
Am I actually using AI or just messing around with it?
The difference shows up in what your business stopped doing, and not in how many tools you have open. Playing has a recognizable shape: you get a good answer, you feel quicker for an hour, and nothing about next Tuesday is different. Real use has a different shape: a job that used to eat an afternoon now takes twenty minutes, and somebody other than you can do it.
Most owners who tell us they are using artificial intelligence across the business are describing a habit that lives entirely in their own head, and it goes on vacation with them.
I use ChatGPT every day but I can’t tell if it’s helping
You cannot tell because you are measuring the tool rather than the job. The tool is impressive every single time, which is exactly why it tells you nothing – impressiveness is not a number you can hold up against last month.
Our advice is to pick one job that comes round at least weekly and time it twice: once your way, and once with the tool taking the first pass. Write down the checking time as well, because checking is where the saving goes.
Do I need an AI strategy, or is that just consultant talk?
You need a shortlist of three jobs, and that takes an hour and no software. Rank them by how often each job comes round, because frequency is what pays you back. Ten minutes saved on a daily job is about forty hours a year, while an hour saved on a monthly one is twelve.
Researchers at the Massachusetts Institute of Technology’s Media Lab reported in 2025 that roughly ninety-five percent of company pilots of these tools showed no measurable effect on profit. Their study looked at companies far larger than yours, and the reason they gave is the useful part: the pilots that paid off went narrow, aiming at one job in one corner of the business.
I automated something and now I spend more time fixing it than it saved
That happens when you point the tool at work where checking the output is the whole job. Some work was always going to be read by a person first: the draft of a quote, a reply to a supplier, the notes from a call. A rough version that needs an edit is a real gain there.
Other work goes straight to a customer or into your books, where a wrong answer costs money and nobody is watching for it. In our experience that second kind eats every hour the first kind saved.
| Playing with the tools | Using them | |
| Name the job | “it helps with all sorts of things” | Monday’s quotes, Thursday’s newsletter |
| Who can do it | only you | anybody on the payroll |
| Where the instructions live | your head and your chat history | written down where staff can find them |
| What you measure | how good the answer was | minutes on that job, checking included |
| If the tool vanished tomorrow | nothing changes | you would notice by Wednesday |
So where do I even start?
Our advice is to start with the job you already dread on a Friday, rather than the tool somebody recommended. Give it two weeks, write down the minutes including checking, and show one other person how it works. If you still cannot name the job or the number, change the job rather than the tool.
Questions owners keep asking
“We bought the AI subscription and nobody on my team uses it”
Nobody uses it because nobody was told which job it was for. A subscription handed out with the words try it is a permission, and everybody already has plenty to do. Pick one job, show one person how you do it, and let that person show the next.
“Is AI worth it for a business with 3 employees?”
Yes, and arguably more than for a big company, because in a business of three there is nobody to hand the boring hour to. The catch is that you have no slack for a bad experiment: one job, one tool, two weeks, and a number at the end.
“How do I stop my staff pasting client info into ChatGPT?”
Tell them plainly what may go in and what may not, in one written paragraph they can find again. Most people paste sensitive material because nobody ever drew the line. Name what stays out, such as customer names, contracts and anything with payment details.
“Everyone says use AI but nobody says for what”
They do not say for what because the answer sits in the jobs only you can see. Look at the work that repeats weekly, involves writing or looking something up, and gets checked by a person anyway. That is your shortlist.
The bottom line is that strategic use of these tools is boring to describe and easy to prove: you can name the job, you can name the number, and somebody other than you can do it. Everything else is practice, which deserves a few weeks of your attention rather than a year.
The team at Sigma Growth Specialists spend a good deal of time helping owners tell those two apart.
