Using AI well does not come down to how often you use it. A busy week of prompts tells you very little about whether the work is better for it.
To see whether you are using AI well, start with something you have actually used it for this week. An email, a proposal or a set of meeting notes will give you more to work with than a general feeling that AI is useful.
After each piece of AI work, I would ask three plain questions.
1. Did it save me time once I had checked it?
The first draft arrives quickly. The time goes on reading it, correcting it and checking the facts. If checking took as long as writing it myself would have, AI did not save time on that task. That is useful to know. Some tasks suit AI and some do not.
Take a follow-up email after a meeting. You give AI a few notes and ask it to turn them into a clear message. The draft takes a minute, but it includes a deadline you never agreed and leaves out the client’s main concern. Finding and fixing those details belongs in your time calculation.
For example, if you normally spend 20 minutes on that email, and preparing the notes, prompting and checking takes 15, you have saved five minutes. If it takes 25, the quick first draft has not saved time overall.
Time is only part of the picture. It might help you organise your thoughts or get started on something you have been putting off. If that is the benefit, name it. You can then decide whether the extra checking is worth it.
2. Would I stand behind this if a client saw it?
Imagine the client reading over your shoulder. Does it say what you agreed? Is there a figure, a date or a promise in there that you did not put in? A smooth paragraph can hide a detail nobody checked. If you would hesitate, it is not ready to send.
Suppose you ask AI to turn your notes into a proposal. Your notes say you will discuss possible next steps. The draft says you will deliver an implementation plan within two weeks. That sounds helpful, but it changes what you are offering.
I would check the proposal against the original notes, paying particular attention to the scope, fees and deadlines. Reading it only for spelling and tone would miss the problem.
There is another check here too: does it sound like something you would say? If an email uses language you would never use with that client, rewrite it. You need to be comfortable with both the detail and the way it is expressed.
3. Did I know what I shared?
Before pasting anything in, I would look for client names, staff matters and figures that should stay inside the business. If you work for an organisation, check which tools it allows for that kind of material. If you cannot say what you put into the tool, start there.
For example, you might want help finding the main themes in meeting notes. Those notes could include a named employee’s health issue or a client’s financial position, even if neither is relevant to the summary you want.
Ask whether AI needs those details for this task. You may be able to use a short, general description instead. Removing names alone may leave enough detail to identify someone, so read the whole example before sharing it.
If you are unsure whether the material is suitable for the tool, pause and check your organisation’s guidance or seek appropriate advice. You can still try the task with a made-up example while that question is answered.
A weekly habit for using AI well
Pick one task you repeat each week. The next three times you use AI for it, answer the questions afterwards and jot down the answers.
For that follow-up email, your note might say: “Five minutes saved after checking. Removed an unagreed deadline. Used only the action points, without confidential discussion.” That is a much clearer record than “AI helped with emails”.
Look for something you can change next time. If you keep removing promises from drafts, tell the tool to use only commitments recorded in your notes, then check that it has followed the instruction. If preparing safe material takes longer than the task itself, doing it yourself may make more sense.
You do not need to assess every tool you have ever tried. Start with one recurring task. A few notes about the result will tell you more about using AI well than counting how many prompts you sent.
Want some help?
If doing this raises more questions, my AI Confidence Check offers individual support for independent professionals with one recurring task. We look at how you use AI now, then work through an agreed example. You receive a written report and a revised process to return to.
If you would like to talk about whether it fits your work, there is a free 15-minute discovery call. No pressure at all. It is a light chat to see if I can help.
Which question would help you check whether you’re using AI well in your next task?
This article is adapted from my LinkedIn article of 21 October 2026.
