Editorial illustration of a Monday planner page with five glowing prompt blocks stacked in priority order against a near-black background, fire orange accents.
AI for ProductivityMay 5, 20264 min read

The 5 Monday Planning Prompts I Run Every Week in Claude

Five copy-paste prompts I run every Monday in Claude or ChatGPT to plan my week. 20 minutes in, you have a ranked plan, drafted replies, and a kill list.

Reeve YewReeve Yew

I run the same five Monday planning prompts in Claude every week before I touch my calendar. Twenty minutes in, I have a prioritised week, drafted replies to every important email, and a short list of things to stop doing. Here is the exact set, copy-paste ready, in the order I run them. All you need is a Claude or ChatGPT account and last week's notes.

What problem does this Monday prompt routine actually solve?

The default Monday for most operators is a slow start. You open your email, you scan Slack, you stare at last week's to-do list, and ninety minutes later you have answered three pings and made no decision. The five prompts below replace that drift with a 20-minute structured pass that produces real outputs: a ranked task list, two or three drafted replies, and a written picture of what a good Friday looks like. (For the wider context on how AI changes the operator's day, see our AI for productivity pillar.) None of it is magic. It is just the same questions a coach would ask you, run by an AI that does not get tired of asking them. The reason it works is consistency. You run the same prompts every week, and the model's outputs compound on top of your earlier weeks because the context block carries forward.

Prompt 1: What did last week tell me?

Here is what I planned for last week and what actually happened: <paste>.
Tell me, in three bullets: (1) the one thing that worked better than expected, (2) the one thing that wasted my time, (3) the pattern you see if you compare this week to the last three weeks I told you about.

This is the only prompt that needs you to bring data. Paste last Monday's plan, last Friday's actual results, and any notes from the week. The point is not the AI's analysis (it is fine, not brilliant). The point is forcing yourself to look at last week before you plan this one. Most operators skip this step and pay for it on Wednesday. I run this prompt before my first email of the day. The "wasted my time" bullet is usually a meeting I should have declined.

Prompt 2: Where will I really spend my time this week?

Here are my commitments and goals for this week: <paste meetings, deadlines, top priorities>.
Estimate where my hours will actually go versus where I want them to go. Flag any single block of time over 90 minutes that is not protected. Suggest 1 to 2 specific meetings I should consider declining or shortening.

This produces a brutally honest week-shape preview. The model is good at spotting the gap between what you said your priorities are and how your calendar is actually loaded. The 90-minute block flag matters: deep work needs at least one of these, and most operators have zero per week without forcing it. I usually decline at least one meeting based on this prompt. The week's hardest decision, made in two minutes.

Prompt 3: Which emails do I owe and what should each one say?

Here is my unanswered email and Slack list: <paste senders + 1-line summaries>.
For each, draft a 3-sentence reply in my voice (direct, warm, no fluff). Group them as: (a) reply now, (b) reply Tuesday, (c) decline or delegate.

This is the prompt that pays for itself in week one. Most of the unanswered messages do not need a paragraph, they need three sentences. The grouping forces a triage decision. The Tuesday bucket is the secret: most messages do not need a Monday answer, and admitting that frees up the morning. I copy the "reply now" drafts into Gmail or Front, edit lightly, send. Total time: 10 minutes for what used to take an hour.

Prompt 4: What is on my list that I should not be doing?

Given my role <one sentence> and this week's goals <paste>, look at my full task list <paste> and tell me which 2 to 3 items I should delegate, defer, or kill outright. Be specific about who or what to delegate to if you can.

The model is surprisingly good at this because it does not have your ego. Items that have been on your list for three weeks are flagged. Items that do not match your role get questioned. I usually kill one task and delegate one per week from this prompt. Over a year that is roughly 50 tasks reclaimed without any heroic effort. Read more on the broader operator workflow in our AI for work coverage.

Prompt 5: What does success look like Friday afternoon?

Based on everything above, write 3 to 5 sentences describing what a great Friday afternoon looks like for me this week. Be specific. Name the deliverables that should exist. Do not include feelings or vague outcomes.

This is the most important prompt. Five sentences. Concrete deliverables. No feelings. Save the output somewhere you will see on Friday morning (a sticky note works fine). When the week wobbles on Wednesday, this is what you steer back toward. The version of you that wrote it on Monday is calmer and clearer than the version that will read it on Friday. That is the trade you are making with the routine.

How to actually run this in 20 minutes

Save your role, your team, and your quarter goals as a Claude Project or ChatGPT Custom GPT instruction once. Every Monday, open that workspace, run prompts 1 to 5 in order, paste outputs into your calendar and task tool, and stop. Do not edit the prompts week to week unless something obviously breaks. The whole pass should fit in one cup of coffee. For more on building this kind of repeatable AI workflow, see our AI how-to coverage. When you have a system that survives the week, join AI Masterminds and share what you changed. We compare prompt sets every Friday.

FAQ

Do I need a paid Claude or ChatGPT plan to run these prompts?

No. The free tiers of both Claude.ai and ChatGPT can run these prompts. You will hit message limits faster on free if you also use them for other work the same day, so the paid tier (around the price of a few coffees per month) pays back quickly. The bigger reason to upgrade is access to Claude Projects or ChatGPT Custom GPTs, which let you save your context once instead of pasting it every Monday. That alone saves five minutes a week.

How long does it actually take to set up the first time?

About 30 minutes the first Monday, 20 minutes every Monday after. Most of the first-time cost is gathering last week's notes and writing your one-paragraph context block (your role, your team, your top three goals for the quarter). Once that block is saved as a Project or Custom GPT instruction, every future Monday starts at prompt 1 with the context already loaded. From week two onwards, the routine is just running the five prompts in order and copying the outputs into your calendar or task tool.

Can I run all five prompts in one chat or should each be separate?

Run them in one chat, in order. The model gets better answers with each prompt because it has the previous prompts as context. Prompt 5 in particular leans heavily on what you said in prompts 1 and 2. The exception is prompt 3 (the email drafts), which I sometimes split into a fresh chat if I want longer drafts without losing context budget on the planning side. If you are using a model with a smaller context window, run prompts 1 and 2 together, then 3 to 5 in a fresh chat with prompt 2's output pasted in.

What if my week never goes to plan and the routine feels pointless?

Run prompt 5 anyway. The Friday-success prompt is the one that compounds. Even if Monday's plan blows up by Wednesday, having written down what 'a good Friday' looks like makes it easier to recover. I did not believe this either until I ran it for a month. The weeks where everything still went sideways still ended better than the weeks I did not write the success picture down. The other four prompts are scaffolding. Prompt 5 is the actual win.

Sources

  1. Claude Projects · Anthropic · June 25, 2024
  2. ChatGPT Custom Instructions · OpenAI

More where this came from

Documentation, not the product.

See all posts →