How will I know if AI is actually working in my business?

It is easy to lose track of whether a tool is working when nobody wrote down how things were before. A month later it feels a bit better, or a bit worse, and the subscription keeps running.

What helps is small, and you do it before you start. Pick one job. Pick one number that describes it in plain terms: the hours you spend on it in a week, how many enquiries go unanswered, how many days a quote sits before it goes out, how many invoices are more than two weeks late. Write the number, the length of time you counted over and the date in your phone's notes. That is your starting line.

Choose a number you can count without a tool. A tally on the back of an envelope beats a dashboard you do not trust. If you cannot count it today, that is worth knowing too: the first job is to start counting, and it costs nothing.

Now change one thing, not three. If you change the reply wording, the quote template and the phone setup in the same week, you will never know which one did anything. Then count the same thing the same way, over the same length of time, and compare like with like: a week against a week, a month against a month.

Choose a review period that fits the job. Missed calls show up within a week. Invoices take time to be paid, so a month is fairer. Anything seasonal needs a comparable stretch of the year. There is no one right number of days.

Decide the rule before you look. Something like: “if the late invoices have not dropped from nine to six after a month, I stop and try something simpler.” Without a rule written in advance, almost any result can be read as a win, especially if you have paid for the tool. This is the same discipline the is it worth it question needs.

Here is a made-up example. A plumber ticks a sheet on the fridge for a week: eleven calls missed, and none of those callers heard from her. She sets up a saved text reply on her phone (see whether a missed call text back is worth it) and ticks the same sheet for another week. Eight calls are missed, but six of those callers answer her text and she rings them back. The outcome she measures is the same both weeks, missed calls that ended with no contact: eleven, then two. The number moved, so she keeps it. If it had stayed near eleven, she would have dropped it and spent nothing.

Be wary of any seller who measures their own tool in their own terms, such as messages sent or calls handled. Those count activity, not whether your business is better off. Your number is jobs booked, hours back or money in. How to tell a real offer from the hype covers what to ask them.

A Three Fixes report is a starting line you can measure from: it names the first steps for your week, and each one comes with a check you can run yourself, so you can judge whether it earned its place.