You used AI during the week. Perhaps it helped you organise a promotion, prepare a checklist or make sense of a difficult task. Before you move on, ask one useful question: what is worth repeating?
A finished answer can feel like progress. The stronger test is whether the work became clearer, easier to use or more reliable after you checked it. A short review helps you keep the useful part of the process and notice where it still needs work.
Set aside about five minutes and choose one ordinary business task. You are reviewing a specific piece of work, so you do not need to judge everything AI can or cannot do. One honest example is enough to start.
1. Look at what you actually used
Open the finished piece of work beside the first AI answer. It might be an appointment reminder, a product description, a meeting agenda or a checklist for packing orders.
Ask what made it into the final version. Did you use the structure? A useful question? A clearer explanation? Or did you rewrite almost everything? “AI helped with my business” is difficult to assess. “It turned my rough notes into a packing checklist that I checked and used” gives you something concrete.
If the answer is still sitting in a chat, record it as a draft. That may be a useful start, but you have not yet tested it in the task it was meant to support. Keep the next step small: check it, try it where appropriate, then decide whether it helped.
2. Count the work after the first answer
Include the time you spent explaining the task, correcting the answer and checking the final result. A quick first draft can still require considerable work before it is ready.
For example, imagine a repair business using AI to prepare a customer handover checklist. The first answer arrives quickly, but it includes services the business does not provide. The owner removes those items, adds the actual collection procedure and checks the wording with a colleague.
The useful result may be a better-organised checklist. Whether it saved time is a separate question. If the owner did not measure the full task, “this gave me a clearer structure” is more accurate than “this saved me an hour”. Next time, they can record the start and finish times, including review, if time saving is what they want to test.
3. Find the correction that matters most
Look for one change you had to make that you would rather not explain again. It might concern your actual service area, the order in which a task happens or a detail the AI assumed without evidence.
Suppose a fictional retailer asked for a comparison of two products. The AI produced a neat summary, but treated an unknown feature as a confirmed benefit. The useful correction is specific: “Use only the supplied product information. Mark missing features as unknown and ask me before making a recommendation.”
Save that correction alongside the task. It gives your next attempt a better starting point. It does not guarantee the same problem cannot happen again, so keep the relevant check in your review.
4. Save a short note you can reuse
You do not need a large prompt collection. A small, understandable record can be easier to maintain. Start with these five lines:
- Task: What was I trying to get done?
- Useful part: What did I actually keep or use?
- Correction: What did I need to change?
- Next time: What instruction or example should I provide earlier?
- Check: What must I verify before using the result?
For the repair business, the note might say: “Prepare a handover checklist. Keep the clear step-by-step structure. Remove services we do not offer. Supply our current collection procedure at the start. Check every service and customer instruction before use.”
Store the note where you can find it when the task comes around again. Keep changing facts, such as current prices and availability, separate from the reusable instructions. Check those facts each time. An old example can help with structure while still containing details that no longer apply.
5. Choose what happens next
Finish with one decision: repeat the approach, adjust it or leave it aside for now. You do not have to keep using AI for a task simply because you tried it once.
If the approach was useful, try it on the next comparable task and see whether it holds up. If it needed too much correction, narrow the task or improve the information you provide. If you cannot judge the result safely or confidently, get appropriate human help before relying on it.
This review can also reveal a useful question to bring to a learning community: “The outline helped, but it kept adding services I do not offer. How would you tighten that instruction?” That gives other people something specific to respond to.
Share the lesson in general terms. Remove customer names, contact details, payment information and confidential material. You can describe the difficulty without posting a screenshot of a private conversation or someone else's records.
A simple way to close the week
Complete three sentences: “I used AI to…”, “I still needed to…” and “Next time I will…”
If you have not used AI this week, choose one small task you would like to try and name the part you are unsure about. That is a useful starting point too.
The aim is to leave with a clearer understanding of your own work. Keep one useful result, one correction and one next step. Then test what you have learned when the task comes around again.