Weekly Review Template

Fifteen minutes, once a week. The only part of any AI system that reliably changes anything.

Why this one thing

Everything else in this course is a technique. This is the mechanism that makes techniques compound. Without it, you use AI the same way in month six as in week one, because nothing ever prompts you to notice what is not working.

The template

Copy this into a note, once a week, same slot.

WEEK OF [DATE]

WHAT I USED AI FOR
[List. Be honest — include the times it did not help.]

WHAT WORKED
[Which task, which approach, what made it work]

WHAT DIDN'T
[Where the output was wrong, useless, or took longer than doing it myself]

TIME SAVED, ROUGHLY
[Hours. Estimate low.]

DID I DO ANYTHING 3+ TIMES?
[Candidates for a saved prompt or an automation]

AUTOMATION CHECK
[Did everything run? Anything misbehaving? Anything I should turn off?]

PROMPT FIXES
[Any prompt I corrected the same way twice? Fix it at the source.]

ONE CHANGE FOR NEXT WEEK
[Just one]

The four questions that matter

"What didn't work?" The section people skip and the only one that improves anything. If it is empty every week, you are not being honest or not trying anything new.

"Did I do anything 3+ times?" This is how your automation backlog builds itself from evidence rather than from a list of ideas you liked.

"Any prompt I corrected the same way twice?" The highest-leverage line here. If you keep telling it "shorter" or "don't use bullet points", that belongs in the prompt or your custom instructions, not in every conversation.

"One change." One. A review that produces five changes produces zero.

The AI-assisted version

If you keep a daily close, let it do the pattern-finding:

Here are my daily notes from this week: [PASTE]

1. What did I use AI for, grouped by type of task?
2. Where did it save the most time?
3. Where did it not help, and is there a pattern?
4. What did I do repeatedly that could be a saved prompt or an automation?
5. What is the single highest-value change for next week?

Be direct. Do not encourage me. If the answer to 4 is nothing, say so.

The monthly extra

Once a month, add three questions:

- Which automations have I not thought about in a month? Are they still running,
  still correct, still worth their cost?
- What am I paying for that I have not used in 30 days?
- What did I automate that I should undo?

That last one is real. Some things are worse automated — the ones where doing it by hand was how you noticed something.

The quarterly extra

- Re-run my test prompts. Has quality changed? (Models update silently.)
- Total AI spend this quarter. Is it in line with the value?
- What can I now do that I could not three months ago?
- What did I plan to build and never did? Is it still worth doing, or should
  I drop it honestly?

Making it survive

Same slot every week. Friday afternoon works for most people — the week is fresh and you are not protecting focus time.

Fifteen minutes, hard stop. A review that takes an hour gets skipped.

Skip twice and it is dead. If you miss two in a row, the honest move is to shorten it rather than resolve to try harder.

Write badly. These notes are for you. Complete sentences are not required and slow you down.

The tracker

Week	Time saved	One change	Did I do it?

The last column is the whole system. A change identified and not made is worth nothing, and after four weeks the pattern in that column tells you something useful about yourself.

What good looks like after three months

• Your prompt library contains things you actually reuse, not things you saved once
• Your automations are fewer than you expected and more useful
• You have turned something off
• You can state your time saved with a number
• You have stopped using AI for at least one thing where it did not help

That last one is a sign of maturity, not failure.