TL;DR
Use a chat assistant (ChatGPT or Claude) to turn last week’s chaos into next week’s structure: three outcomes, one anchor task per weekday, an honest “drop” list, and a three-question Friday review. Set it up once as a saved Project so you never re-paste your goals. Keep the decisions yours — AI is the editor, not the boss. Budget 25-40 minutes. The single most-skipped step, the Friday review, is the one that actually changes the next plan.
Why Sunday is where the week is won or lost
Most weeks fail not on Wednesday but on Sunday — when you sat down expecting to plan and ended up scrolling. The point of a weekly planning ritual is to walk into Monday with a small number of clearly chosen outcomes and a quiet conscience about everything you decided not to do.
This is the part productivity systems agree on and most people skip. In David Allen’s Getting Things Done, the weekly review is the keystone habit, and it is also the one most often abandoned. The reason to use AI here is not novelty — it is that an assistant will ask the question a good friend would (“you wrote this same goal three weeks ago; what changed?”) without the friction of scheduling that friend, and without judgement.
When AI helps — and when it does not
AI helps you turn last week’s chaos into next week’s structure: clustering tasks, exposing duplication, surfacing what kept sliding. It does not know your energy levels, your kid’s swimming schedule, or that you secretly hate the 6 a.m. gym slot you keep scheduling. Treat AI as the planner’s editor, not the planner’s boss. The final say on what makes the week stays with you.
| AI is good at | You stay in charge of |
|---|---|
| Spotting tasks that repeat week after week | Whether a repeated task still matters |
| Clustering 30 loose to-dos into 3-5 themes | Which theme is actually this week’s priority |
| Flagging an over-booked week (no buffer left) | Trading off family vs. deadline pressure |
| Drafting a Friday review from your plan + reality | Being honest in the answers |
| Suggesting a time-block layout | When you actually have brainpower |
Pick the tool, then set it up once
You do not need a dedicated planning app for this. The fastest path is the assistant you already pay for, set up as a reusable workspace.
| Approach | Cost (June 2026) | Best for | Setup |
|---|---|---|---|
| ChatGPT Projects | Free tier works; Plus $20/mo for higher limits | Most people; pin goals + last week’s review as project files | Create a Project, add custom instructions, upload your goals doc |
| Claude Projects | Pro $20/mo ($17 annual; bundles Claude Code + Cowork) | Long context — paste a whole quarter of notes (200K-token project knowledge) | Same idea; project instructions act as a permanent system prompt |
| Gemini (in Google AI Pro) | $19.99/mo, includes Workspace | If your tasks live in Google Calendar / Docs | Use Gems for a saved planning persona |
| Dedicated AI calendar (Sunsama, Reclaim, Motion) | Sunsama from $16, Reclaim free/$12, Motion ~$19/seat | People who want the plan auto-placed onto a real calendar | Connect your calendar; let it time-block |
The Projects feature in both ChatGPT and Claude is the unlock for recurring planning: you save your quarterly goals and formatting rules once as project instructions, and every new Sunday chat starts with full context instead of a blank box. As of June 2026, ChatGPT Projects keep memory scoped to that project (it does not bleed into your main chats), and Claude project knowledge holds about 200K tokens — roughly 500 pages — so you can keep a running log of past weeks.
If you genuinely want the plan dropped onto a calendar with auto-scheduling, that is the line where a dedicated tool earns its price. The chat assistant designs the week; an AI calendar places the blocks. See AI calendar planning tutorial for that hand-off.
What to feed the AI
- Quarterly goals (3 max) — anything more is wishful thinking
- Hard constraints for the upcoming week (travel, deadlines, on-call, family)
- Last week’s wins and drops — what you finished, and what you postponed for the third time
- Your real available focus hours (not your calendar — the hours you actually have brainpower)
- One “no” you would like the AI to enforce — e.g., “no meetings before 10 a.m.”
In a Project, the first three rarely change week to week, so put them in the project instructions and only paste the last two each Sunday.
Copy-ready prompt
Plan my coming week.
Quarterly goals: [up to 3]
Constraints next week: [list]
Last week wins: [list]
Last week drops / postponed: [list]
Available focus hours per day: [numbers, Mon-Sun]
Hard "no" I want you to enforce: [rule]
Return:
1. Top 3 outcomes for the week (each tied to a goal)
2. One daily anchor task per weekday — the thing that, if done, makes the day count
3. What to explicitly drop or defer, with reason
4. One "wild card" creative or learning slot (90 min)
5. Friday review checklist (3 questions max)
6. A "you wrote this last week too" flag — name anything that keeps showing upVariant for ICs in a busy quarter: “Same plan but assume Wednesday is a no-meetings deep-work day and protect it.”
For the Friday review, reuse the same chat: paste what actually happened and ask, “Compared to Sunday’s plan, what kept sliding, and why?” Because the plan is already in context, the model can name the gap instead of guessing at it.
Recommended output structure
A one-page week-at-a-glance: outcomes at the top, daily anchors in a 5-row table (Mon-Fri), drops in a strikethrough list. The Friday review should be three questions — not a retrospective template. If you want the plan to live in software rather than a doc, ask for the daily anchors as a checklist you can paste into Todoist, Things, or your AI calendar.
How to check the output is usable
- The top 3 outcomes are different from last week’s top 3, OR they explicitly say “carried over because…”
- The daily anchors take less than 2 hours each — anchors are not entire days
- “Drop / defer” actually has items in it. A plan that drops nothing is not a plan
- Total committed hours leave at least 20% buffer for the inevitable surprise
- The Friday review questions are answerable in 5 minutes, not 30
Common mistakes
- Over-scheduling — every hour booked, no buffer. Knowledge workers already lose a large share of focus to interruptions and task-switching; a packed plan guarantees the first surprise breaks it
- Letting AI re-rank your goals without permission. AI can question priorities; it should not silently reshuffle them
- Putting the gym at 6 a.m. because the AI suggested it. The AI does not know you
- Re-pasting your whole context every week instead of saving it in a Project — you will eventually skip the ritual because the setup is annoying
- No Friday review — without it, every week’s plan repeats the same drops
- Mixing the weekly plan with a project plan. A weekly plan is not a Gantt chart
FAQ
- Sunday or Friday for planning? Either works, but pick one and protect it. Sunday gives you a clearer head; Friday gives you a softer Monday. The format matters less than doing it the same time every week.
- How long should this take? 25-40 minutes once it is a habit. The first few reviews run 60-90 minutes while you build the muscle; that is normal. If your weekly plan still takes 90 minutes after a month, you are planning instead of doing.
- ChatGPT or Claude for this? Either is fine on the $20/mo tier. ChatGPT Projects are slightly more polished for casual use; Claude’s larger project knowledge (~200K tokens) is better if you keep a long log of past weeks. The free tiers of both can run this prompt — you just lose saved Projects and hit usage limits faster.
- Do I need a paid AI calendar like Motion or Sunsama? No, not to plan. A dedicated calendar (Sunsama from $16/mo, Reclaim has a free tier, Motion ~$19/seat as of June 2026) earns its price only when you want the plan auto-placed and rescheduled on your real calendar. For deciding what the week is, a chat assistant is enough.
- Can AI handle the Friday review too? Yes — feed it the Sunday plan plus what actually happened, in the same chat. Ask: “What kept sliding and why?” That feedback loop is the whole point; a plan you never review just repeats the same drops.
Related
- Weekly planning prompts — alternative ritual phrasings
- Task prioritisation with AI — break weekly outcomes into daily priority order
- Habit tracker with AI — make the Sunday ritual itself a habit
- Decluttering plan — same prioritisation logic for the home
- AI calendar planning tutorial — push the weekly plan into your actual calendar
External references: David Allen’s Weekly Review (GTD) and OpenAI’s Projects in ChatGPT.
Tags: #Productivity #Workflow