The weekly time review is straightforward in theory: look at how you spent your time, learn something, change something.
In practice, most attempts stall at the first step — gathering the data — or at the last step — actually making a change. This guide walks through the full process concretely, with the specific prompts and formats that make each stage work.
Set aside 30 minutes this Friday. Here’s exactly what to do.
What Do You Need Before You Start?
You need three things: your calendar for the past week, a rough sense of what your intended priorities were, and an honest memory (or better, a note) of what actually got done.
Optional but useful: any time-tracking data from tools like Toggl, Clockify, or Harvest; a task list showing what you planned versus completed; and a brief subjective note on energy — which days felt good, which felt scattered.
If you have none of this except your calendar, that’s enough to start.
Step 1: Categorize Your Calendar (5 minutes)
Open your calendar and look at the past five working days. Assign every block to one of these four categories:
- Deep work: Focused, cognitively demanding work — writing, building, analysis, strategy. Work that requires your full attention and that no one else can do for you.
- Meetings and calls: Any synchronous communication, whether external (clients, partners) or internal.
- Admin: Email, Slack, scheduling, expense reports, any task that’s reactive and interruptible.
- Personal/buffer: Breaks, lunch, commute, any non-work time in the workday.
Estimate hours per category. You don’t need precision — estimates rounded to the nearest half-hour are fine.
Then add two lines:
- What were your intended top-three priorities this week?
- What actually got completed?
Your input block should look something like this:
Week of [date]
Time by category:
- Deep work: [X] hours
- Meetings/calls: [X] hours
- Admin: [X] hours
- Personal/buffer: [X] hours
Intended priorities:
1. [Priority 1]
2. [Priority 2]
3. [Priority 3]
What I actually completed:
- [Item 1 — done/partial/not started]
- [Item 2 — done/partial/not started]
- [Item 3 — done/partial/not started]
Energy note: [Which days felt good? Which felt scattered or resistant?]This takes five minutes if you do it at end-of-week while the days are still fresh. It takes fifteen minutes on Monday morning when you’re trying to reconstruct it from memory.
Step 2: Run the AI Analysis (10 minutes)
Open your AI tool of choice — Claude, ChatGPT, or any capable language model. Paste this prompt with your data block filled in:
I'm running my weekly time review. Below is my time data for the past week. I want a structured retrospective with three specific outputs — nothing more:
1. ONE WIN: The most meaningful pattern or behavior in how I spent my time this week. Not just a task I completed — a habit, decision, or allocation worth reinforcing.
2. ONE LEAK: The single biggest gap between how I spent my time and what I said my priorities were. Be specific: name the category, estimate the hours, and explain the cost.
3. ONE SHIFT: One concrete change I should make to next week's schedule. Not a general principle. A specific reallocation — e.g., "Move your Monday 9am meeting to Thursday afternoon and use that slot for deep work on [priority]."
Base everything on my stated priorities and the data I've provided. Don't give me general productivity advice.
[Your data block]The three-output constraint is intentional. An unconstrained prompt produces a list of observations — which is interesting but not actionable. The win/leak/shift structure forces the AI to prioritize and commit to specific claims, which gives you something concrete to react to.
Step 3: Read and React (7 minutes)
Read the output once without immediately critiquing it.
Then go through each output and ask yourself:
On the win: Does this feel true? Is there something the AI identified that you hadn’t consciously noticed? If the win doesn’t resonate, what would you name instead?
On the leak: This is the most important output. The natural response is mild defensiveness — the AI is naming something that didn’t go well, and the impulse is to explain why it was unavoidable. That impulse is worth examining. If you want to push back, ask: “Am I pushing back because the AI is wrong, or because I don’t want it to be right?”
On the shift: Is this something you would actually implement? If not, why not — and what would you actually change?
If the output prompts a useful reaction but leaves something important unnamed, follow up:
You identified [leak] as my main leak. I think there's something you missed: [describe what you think the actual issue is]. Does that change your analysis of the shift I should make?The back-and-forth is often where the most useful insight lives.
Step 4: Write Your Three Sentences (3 minutes)
Before you close the AI and move on, write three sentences — one for each output — in your own words. Not the AI’s words. Yours.
This matters because the act of restating forces genuine understanding. If you can’t state the win in one sentence, you haven’t processed it. If you can’t describe the leak without hedging, you’re still rationalizing.
Your log entry should be:
Week of [date]:
Win — [one sentence in your own words]
Leak — [one sentence in your own words]
Shift — [one sentence: the specific change you're committing to for next week]Save this somewhere you’ll see it. A Notion page, a note in your calendar, a dedicated document. The location doesn’t matter. What matters is that it exists and you’ll encounter it.
Step 5: Schedule the Shift (5 minutes)
Here’s where most reviews fail: the shift stays in the log and doesn’t make it into the calendar.
Open your calendar for next week. Find the slot your shift refers to and make the change now, while the reasoning is fresh. If your shift is “protect Tuesday morning for deep work,” block it now. If it’s “move the standing Monday 4pm sync to Thursday,” send the reschedule request now.
If the shift requires a conversation — “I need to reduce my meeting load this week” — write the first sentence of that email now. Don’t put it on a to-do list. Start it.
The review produces value only when it changes behavior. The shift is the mechanism. Scheduling the shift while you’re still in the review is the difference between a reflection that compounds and one that evaporates.
What Makes This Review Stick After Week One
A first review is easy. The second one is a little harder. The fourth one is where most people lapse.
Two habits make the difference.
First: the calendar block is non-negotiable. Treat Friday 4–4:30pm (or your equivalent) as a meeting you cannot cancel without rescheduling. When something conflicts with it, move the other thing, not the review. The moment you treat it as flexible, it becomes optional. Optional reviews don’t happen consistently.
Second: lower the bar in hard weeks. A minimum viable review is three sentences — win, leak, shift — written in five minutes without AI. This is enough to maintain continuity. The goal is 48 reviews in 52 weeks, not 30-minute deep dives every Friday. Imperfect consistency beats perfect sporadic analysis.
What the Data Tells You After Eight Weeks
After eight consecutive reviews, your log becomes its own data source.
Look at eight weeks of leaks. Is the same category appearing repeatedly — meetings, admin, reactive work? That pattern is not an accident. It’s a structural problem in how your week is organized, not a series of unlucky weeks.
Look at eight weeks of shifts. How many did you actually implement? A shift that appears three times in a row without implementation is telling you something: either the shift is right and you’re avoiding it, or the AI is identifying a symptom but not the cause.
This retrospective-on-the-retrospective is what separates the practitioners who make sustained progress from those who improve and regress in cycles. The weekly review creates data. The eight-week review teaches you what to do with it.
Your action: Open your calendar right now and block this Friday from 4:00–4:30pm as a recurring event titled “Weekly Time Review.” When that block arrives, come back to Step 1 above.
The full framework behind this process — including the research on why structured reflection works and how to interpret patterns over time — is in the Complete Guide to Weekly Time Review with AI.
Frequently Asked Questions
-
Do I need a time-tracking app to do this review?
No. Your calendar alone is enough to start. Most knowledge workers have enough calendar data — meetings, blocks, recurring events — to generate a useful review. A rough daily note estimating unscheduled work in 3–4 categories adds precision, but it's optional. Start with what you have and add more data sources as the habit develops.
-
How specific does my time data need to be?
Specific enough to identify patterns — not precise to the minute. Estimates rounded to the nearest half-hour are adequate. The AI is looking for relative proportions (how much of your week went to meetings vs. deep work vs. admin) and alignment with your stated priorities. Rough data that's complete is more useful than precise data that's missing categories.
-
What if the AI's analysis feels wrong?
Push back directly. Tell the AI why you disagree and what you think is missing. The most productive reviews often involve a short back-and-forth where the initial analysis prompts a reaction, you provide additional context, and the AI refines its read. The goal is your honest assessment, with the AI as a thought partner, not a verdict-giver.