AI Guides › Playbooks

The Ninety-Day Quarter Card: A Five-Pass Review in One Claude Chat

By Nigel Guy · 7 min read

Most people review a quarter by asking an AI "how did my last three months go?" and pasting in a wall of half-remembered notes. What comes back is warm, balanced and forgettable, because you gave it vibes and it gave you vibes. Three weeks later nothing in your calendar has changed.

The rule: a review is finished only when it produces three priorities, a weekly check-in and a list of things you are dropping, all written down where next quarter's you can find them.

The mechanism: the Quarter Card

The Quarter Card is one page you build across five passes in a single chat. Each pass has a job, and you do not move on until Claude has produced that pass's output. One chat matters because each pass builds on the last; if you split them, Claude starts each one cold.

Pass Job Output on the card
1. Dump Get the raw quarter out of your head A plain list of events, decisions, wins, misses
2. Score Sort what happened by what it cost and what it gave Keep / Change / Stop columns
3. Goals Test your goals against the evidence Goals marked "backed by what I did" or "stated only"
4. Choose Cut to three priorities Three priorities, each with a done-test
5. Weekly Turn the priorities into a 15-minute check-in A repeatable weekly question set

Before you run it

You need about 60 to 90 minutes and a calendar, a notes app or a bank statement to jog your memory. Claude's free plan includes chat and memory; Projects are listed on the Pro plan, which was $20 a month ($17 on annual billing) at time of writing, about £15 at time of writing, so check the £ price at checkout. Support documentation also says free users can create projects (up to five), so the plan matters less than you might think. If you want the review to persist, make a Project: go to claude.ai/projects, choose "+ New Project", name it "Quarterly review", and add the finished card to the project knowledge with the "+" button. Each project keeps its own memory space, which is useful for keeping a personal review out of your work chats.

Pass 1: Dump

Do this before showing Claude any goals, so the goals do not shape what you remember.

You are a calm, practical review partner. I am doing a 90-day review for [START_DATE] to [END_DATE].
I will paste raw material in several messages: [CALENDAR_NOTES], [JOURNAL_OR_NOTES], [MONEY_OR_TIME_RECORDS].
Do not analyse yet. After each message, reply only with "Got it" and one line naming anything you need clarified.
When I type DONE, produce a chronological list of what happened, grouped by: work, money, health, relationships, learning, other.
Flag anything that looks like a gap in my notes rather than filling it with a guess.
Before replying, check you have added nothing I did not tell you.

Fill in the dates and paste what you have. The "gap" instruction is the guardrail: Claude should say "nothing recorded for June" instead of inventing a plausible June.

Pass 2: Score

Using only the list above, sort each item into three columns:
KEEP (gave me energy or results worth repeating), CHANGE (worth doing but done badly), STOP (cost more than it gave).
For each item, give a one-line reason that points to something I actually said.
Mark any item you are unsure about with "?" and ask me about it. Do not decide for me.
Finish with the three patterns you see across the columns, and tell me which one rests on the thinnest evidence.

Answer the question marks. This is where you correct Claude, not the other way round.

Pass 3: Goals

Now bring in your goals: [GOALS_FROM_LAST_QUARTER_OR_YEAR].

Here are the goals I set: [GOALS].
For each goal, check it against the scored list above and label it:
BACKED (my quarter shows real effort toward it), STATED ONLY (I said it matters but the record shows little), or CONFLICTED (something I did works against it).
Quote the item from my list that justifies each label.
Then ask me up to five questions about the STATED ONLY goals, such as whether I still want them. Wait for my answers.

This pass is the honest one. A goal with no trace in your calendar is a wish, and the useful question is whether you still want it.

Pass 4: Choose

From everything above, propose exactly three priorities for the next 90 days, starting [NEXT_START_DATE].
For each: one sentence on what it is, why it beat the alternatives, a "done looks like" test I could check by looking at a calendar or a document, and the single biggest thing likely to get in its way, based on my record.
Also list what I am explicitly NOT doing this quarter, including at least two goals from above I should drop or park.
Check each priority against my stated limits: [HOURS_PER_WEEK_AVAILABLE], [FIXED_COMMITMENTS]. If three will not fit, say so and propose two.

If Claude hands you five priorities, push back. Three is a constraint, not a suggestion.

Pass 5: Weekly

Build a 15-minute weekly check-in for me, to run on [DAY_AND_TIME].
Use no more than five questions. Include: one about progress on each priority against its done-test, one about what slipped, and one about what I will do first next week.
Then produce the full Quarter Card as one page: my three priorities with done-tests, the not-doing list, the weekly questions and the biggest risk for each priority.
Use headings and short lines, so I can paste it into a note.

Save the card to your project knowledge, or paste it somewhere you will actually see on a Monday.

A worked example (hypothetical)

Imagine Sam, a part-time designer. The dump shows two client projects, a lot of evening freelance work and no mention of the fitness goal. The scored list puts evening freelance under STOP: it paid, but the notes say Sam was tired every Thursday. In the goal pass, "launch a portfolio site" comes back STATED ONLY, because nothing in the quarter touched it. Sam says it still matters. Pass 4 gives three priorities: ship the portfolio site (done-test: live URL), cap evening work at two nights, and keep one fixed exercise slot. The not-doing list parks a newsletter idea. The weekly check-in runs on Sunday evenings.

The trap

Treating Claude's tidy summary as the truth. It only sees what you pasted, so a flattering or gloomy dump produces a flattering or gloomy review. The fix is the evidence rule in each prompt: quote my words, flag gaps, ask when unsure. Memory can also quietly fill in context from earlier chats. It is on by default for Free, Pro and Max, and you can pause it, reset it, or switch it off for a single chat before your first message via the "+" menu. Decide which you want before you start.

What to skip

Guardrails

Sources

All 751 AI guides · JulieMango plans from £17/mo