A busy week and a good week feel exactly the same on a Friday night. That is the problem this guide fixes.
Once a week, for 20 minutes, you tell the AI what you planned to do, what you actually did, and any numbers you track. It compares the 2 and tells you the truth: what moved, what did not, and what you have now put off 3 weeks running.
Then it gives you 3 priorities for next week, based on what actually happened rather than on whatever felt most urgent on Thursday.
That is the whole thing. Below is the set-up, the prompt to paste, and what to do when the answer stings a bit.
Set it up so it accumulates
Do this in a project called Weekly review, for one reason: the reviews then pile up in one place, and the pile is worth more than any single review.
Project instructions:
This project is my weekly business review. I run [what you do]. [Solo, or with staff, or with a partner.] My realistic working hours are [be honest, including if it is evenings].
Be direct and do not cheerlead. If I planned 5 things and did one, say that. If something has slipped repeatedly, name it every time until it moves or I delete it. Do not tell me I am doing well unless the numbers say so.
Australian English. Reviews stay under 300 words.
The review prompt
Weekly review.
What I planned this week: [paste or summarise]. What actually happened: [wins, misses, surprises, interruptions]. Numbers, if I track any: [sales, enquiries, jobs, hours, whatever you have]. Last week's review: [paste it, or say none].
Give me: 1. What genuinely moved forward. Plain terms, no inflation. 2. Where the plan and the reality split, and your best read on why. 3. Anything that has now slipped more than once. Be blunt about it. 4. One thing to stop doing or drop entirely. 5. Next week's top 3, each with a first step small enough to start Monday morning.
No cheerleading. Under 300 words.
Point 4 is the one people skip and it is the most valuable. A review that only adds is a review that grows your list every week until the whole exercise collapses under itself.
Constructed example, a 2-person landscaping business.
Moved forward. The Nguyen job finished and invoiced, 3 days early. 2 quotes out.
Plan against reality. You planned 4 quotes and sent 2. The pattern is that quoting happens at night after site work, and 2 nights this week you were too knackered. That is not a discipline problem, it is a scheduling one. Quote in the truck at lunch or accept 2 a week.
Slipped more than once. The insurance review, 3 weeks running. Either book it now or take it off the list, because carrying it is costing you more than doing it.
Stop doing. Driving to quote jobs under $2,000. 2 hours round trip for a job you may not win.
Next week's 3. 1) Insurance, Monday 8am, one phone call. 2) Quote the Ballarat job, Tuesday lunch, in the truck. 3) Decide on the second trailer before Friday, the Ballarat job depends on it.
The part that makes it compound
One review is a useful Friday. 12 reviews is something you cannot get any other way: an accurate picture of how you operate, as opposed to how you think you do.
Once a month, in the same project:
Read back over the last 4 reviews. What has slipped more than twice, what keeps appearing in the "stop doing" list and never stops, and what is the pattern in the weeks that went well against the weeks that did not? Be specific and do not be kind about it.
Once a quarter:
Read every review in this project. What has actually changed in this business over the quarter, what has stayed exactly the same despite appearing in most reviews, and what would you say is the one bottleneck if you had to name a single one?
That quarterly one is the closest thing a small business gets to a board meeting, and it costs 20 minutes and nothing else.
Make this prompt yours
Feed it your real week and this becomes your review prompt. Nothing you type leaves this page.
Make it survive
Same time every week. Friday afternoon or Sunday evening, whichever one you will really do. A review that happens when you remember does not happen.
Do it on the bad weeks especially. The bad weeks carry the information. The temptation to skip is strongest exactly when the review would tell you something.
Small numbers still count. 3 sales and 2 enquiries is data. Small numbers reviewed weekly beat a dashboard you look at twice a year.
20 minutes, hard stop. If it is taking an hour it has turned into planning, and planning is a different activity that expands to fill whatever you give it.
What a review cannot tell you
It only sees what you tell it. A review built on a rosy account of the week returns a rosy review. The uncomfortable line in "what actually happened" is the one doing the work.
It cannot judge whether your priorities are right. It can tell you that you planned 4 quotes and sent 2. It cannot tell you whether quoting is the right thing to be spending nights on. That one is yours.
Blunt is a setting, not insight. Asking it not to cheerlead gets you a more direct tone. It does not give it access to anything you did not paste.
It is not a substitute for the numbers. If you do not know your actual margin, 12 weeks of reviews will not find it. The review works on what you measure, so measuring something is the prerequisite.
Run it this Friday
Run it once this Friday, with no last week's review to paste. The first one is the weakest, because there is nothing to compare against. Run it 3 weeks in a row and then judge it.