A good AI prompt for weekly planning gives the model your real constraints before it asks for a plan: the hours you actually have free, your fixed commitments, your deadlines, the one or two goals the week should move, and an honest task list. With that input, the model can sort tasks by impact and urgency, fit them into the available time, and tell you what does not fit, which is the part most plans skip. The templates below cover the full cycle: capture, prioritize, fit to capacity, and a short Friday check.
Each template is written to be pasted into any chat model (Claude, ChatGPT, Gemini) with the bracketed fields filled in. Under each one is a note on why its parts are there, so you can adapt it without breaking it.
Why “help me plan my week” gives you a generic plan
A planning prompt without constraints produces advice, not a plan. The model does not know that Tuesday is full of meetings, that the report is due Thursday, or that you have about four hours of real focus time a day. So it fills the gap with defaults: “start with your most important task”, “batch your email”, “schedule breaks”.
The second failure is subtler. Even with a full task list, a model will usually try to fit everything in, because you asked it to plan the week, not to cut it. People already do this to themselves. Psychologists call it the planning fallacy: the tendency to predict that a task will take less time than similar tasks took in the past, even while remembering that those tasks ran long. Roger Buehler, one of the researchers who studied it, summarizes the evidence and the countermeasures in an SPSP article on the planning fallacy. A planning prompt should therefore make the model check your estimates against capacity and explicitly name what is left out.
Template 1: the weekly input sheet
The input sheet is the context block that every other template in this guide reuses. Fill it in once at the start of the week (it takes about ten minutes) and paste it at the top of each planning prompt.
<week>
Week of: [YYYY-MM-DD]
Role: [one line, e.g. "infrastructure engineer on a 6-person platform team"]
Goals this week moves (max 2):
1. [e.g. "Ship the backup-restore runbook before the audit on the 30th"]
2. [...]
Fixed commitments (cannot move):
- Mon: [e.g. 10:00-11:00 team sync]
- Tue: [...]
- ...
Realistic focus capacity: about [N] hours per day of uninterrupted work
(be honest; count what you actually got last week, not what you hope for).
Hard deadlines:
- [task] due [day]
- ...
Tasks (one per line: task | rough estimate in hours | who is waiting on it, if anyone):
- [...]
Energy notes: [e.g. "best focus 8-11am; Friday afternoons are low"]
</week>Why each field is there
- Max two goals. Prioritization needs something to prioritize against. With five goals, everything matches one of them and the ranking means nothing.
- Realistic focus capacity. This is the number that keeps the plan honest. Using last week’s actual figure is a simple version of what Buehler calls taking the “outside view”: basing the estimate on what similar weeks really looked like.
- Estimates per task. Rough is fine. Without them, the model cannot tell you the week is overloaded.
- Who is waiting. A task that blocks a colleague has a cost that does not show up in its own importance. This field lets the model surface it.
- Energy notes. They let the plan put demanding work where it has the best chance of getting done.
If your task list is still a mess of half-formed notes, clean it first with the prompts for turning a brain dump into a prioritized to-do list, then paste the result into the Tasks field.
Template 2: prioritize by impact and urgency
Impact-versus-urgency prioritization means scoring each task twice, once for how much it moves your goals and once for how soon it must happen, and keeping the two scores separate. The idea is usually credited to Dwight Eisenhower, who in a 1954 speech described two kinds of problems, the urgent and the important; according to Quote Investigator’s research on the quote, he attributed the thought to an unnamed former college president. The four-quadrant grid built on it is now called the Eisenhower matrix.
[paste the <week> block]
Score every task in the list on two separate scales:
- Impact (1-5): how much finishing it this week moves Goal 1 or Goal 2. A task that moves neither goal scores 1 or 2 at most, however busy it feels.
- Urgency (1-5): how bad it is if it slips to next week. 5 = hard deadline this week or someone is blocked; 1 = no consequence.
Then place each task in one quadrant:
- Do first: Impact 4-5 and Urgency 4-5
- Schedule: Impact 4-5, Urgency 1-3
- Delegate, shrink or batch: Impact 1-3, Urgency 4-5
- Drop or defer: Impact 1-3, Urgency 1-3
Output a table: Task | Impact | Urgency | Quadrant | One-line reason.
After the table:
1. Name the "Schedule" task most likely to be neglected this week, and why.
2. For each "Delegate, shrink or batch" task, suggest the smallest version that satisfies whoever is waiting.
3. List the "Drop or defer" tasks and ask me to confirm each one rather than deciding for me.The key line is the cap on impact for tasks that do not touch either goal. Without it, models (and people) give high impact to anything that sounds work-related. The neglected-task question targets the quadrant that matters most for longer-term progress: important work with no deadline, which loses to urgent work every single week unless it is protected. The final instruction keeps decisions about dropping work with you, since the model cannot know the politics behind a “low-impact” request from your manager.
Template 3: fit the plan to your actual capacity
Capacity fitting is the step that turns a ranked list into a week you can actually finish. It adds up estimates, compares them to the hours available, and makes the overflow explicit.
Using the prioritized table above and the <week> block:
1. Calculate available focus hours: [N] per day minus time taken by fixed commitments on each day. Show the per-day figure.
2. Add a buffer: inflate every estimate by [30]% for unplanned work and underestimation. State the adjusted total.
3. Allocate tasks to days in this order: "Do first" tasks before their deadlines, then "Schedule" tasks in my best-energy slots, then "Delegate/shrink/batch" tasks in low-energy slots.
4. Stop allocating when a day's hours are full. Do not overbook any day.
5. Output:
- A day-by-day plan (Mon-Fri): task, hours, and whether it goes in a high- or low-energy slot.
- A "Does not fit" list with the hours it would need.
- For each "Does not fit" item: the consequence of deferring it, and who should be told.
6. If any hard deadline cannot be met within capacity, put that warning first, before the plan.The buffer percentage is yours to set; there is no universal right number. Start with whatever gap you saw between last week’s plan and last week’s reality, and adjust. The “who should be told” line matters more than it looks: telling someone on Monday that their request moves to next week is easy, while telling them on Friday is not. If you want the plan expressed as calendar blocks rather than a task list per day, the prompts for a weekly time-blocking plan take this output as input.
A worked example
The following example is invented for illustration. An engineer has 4 focus hours a day, a team sync Monday morning and a two-hour architecture review Wednesday afternoon. Goal 1 is to finish a backup-restore runbook before an audit the following week.
| Task | Estimate | Impact | Urgency | Quadrant |
|---|---|---|---|---|
| Write and test backup-restore runbook | 8h | 5 | 4 | Do first |
| Fix failing nightly pipeline (team blocked) | 3h | 3 | 5 | Delegate, shrink or batch |
| Draft proposal for log retention change | 4h | 4 | 2 | Schedule |
| Review three pull requests | 2h | 2 | 4 | Delegate, shrink or batch |
| Update personal dotfiles | 2h | 1 | 1 | Drop or defer |
| Vendor evaluation spreadsheet | 5h | 3 | 3 | Drop or defer |
The task estimates add up to 24 hours. Capacity is 20 focus hours minus roughly 1 hour for Monday’s sync and 2 for Wednesday’s review, so about 17. With a 30% buffer, the full list needs about 31 hours, nearly double what is available. The runbook, pipeline fix and PR reviews alone come to 13 hours, or about 17 with the buffer, which already fills the week. So the capacity step would put the pipeline fix first thing Monday because a team is blocked, the runbook in the morning slots Monday to Thursday, and PR reviews in low-energy afternoon slots. The log-retention proposal, the vendor spreadsheet and the dotfiles land in “Does not fit”, with a note to tell whoever asked for the vendor evaluation that it moves to next week.
That outcome is uncomfortable, and it is the useful part. The proposal is exactly the important-but-not-urgent task Template 2 warned about. A sensible response is to shrink it rather than lose it: ask the model for a one-hour version (an outline and the key question to answer) and swap it in for one PR review that a teammate can take.
Notice that the pipeline fix scored only 3 on impact but still goes first: blocking a team is a cost the urgency score captures. That is the reason for keeping the two scores separate instead of combining them into a single number too early.
Template 4: the mid-week replan
A mid-week replan is a short prompt for when something new arrives, rather than starting over. Most weekly plans break on Tuesday or Wednesday; this keeps the replan to two minutes.
Here is my plan for the rest of the week:
[paste remaining days from the plan]
New item: [task] | estimate [h] | requested by [who] | needed by [when]
Already slipped: [anything not done that was planned for earlier days]
Score the new item for Impact and Urgency against the same goals.
Then propose the minimum change to the plan:
- What moves to make room, and to which day or to "Does not fit".
- Whether anything should be renegotiated with the requester instead (e.g. a smaller version, a later date).
Do not rebuild the whole week. Keep the output under 120 words.The word limit and “minimum change” instruction stop the model from producing a fresh, beautiful plan that you will abandon again on Thursday.
Template 5: the five-minute Friday check
The Friday check compares the plan with what actually happened and produces one adjustment for next week, which is what makes the next input sheet more accurate. It is deliberately short; a full weekly review with reflection on goals is a different exercise, covered in the prompts for running a weekly review.
Monday's plan:
[paste the day-by-day plan]
What actually got done (task | actual hours):
[...]
What came in unplanned:
[...]
In under 150 words:
1. Planned hours vs actual focus hours available this week.
2. Which estimates were furthest off, and in which direction.
3. Which quadrant lost the most time.
4. One specific change to next week's input sheet (e.g. "use 3.5 focus hours a day, not 4", "add 2h for PR reviews as a fixed block").
Do not give general productivity advice.Over a few weeks, point 4 calibrates the two numbers that drive everything else: your real focus capacity and your typical underestimation. That calibration is what turns these templates from a one-off exercise into a planning system.
Mistakes that make AI weekly plans fail
- Too many goals. If the week has five goals, prioritization collapses. Pick two; everything else is maintenance.
- Vague tasks. “Work on migration” cannot be estimated or finished. “Write rollback steps for the database migration” can. Ask the model to rewrite vague tasks as concrete ones before scoring them.
- Optimistic capacity. Entering 8 focus hours a day when meetings and messages leave you 4 guarantees an overloaded plan.
- Accepting the first plan. Push back: “Which of these would you cut if I had one day less?” is a quick way to test whether the ranking holds.
- Sharing more calendar detail than needed. The model needs time blocks and labels, not attendee names, meeting content or client details. Generalize sensitive entries (“client call”) before pasting.
- Letting the model decide what to drop. It does not know which requests carry weight in your organization. Template 2 asks it to propose drops, not make them.
Making it a weekly routine
The routine is simple: fill the input sheet and run Templates 2 and 3 on Monday morning, use Template 4 when something new lands, and run Template 5 on Friday. Save the five templates in a notes app or as a project in your chat tool so the input sheet carries over and only the changed fields need editing.
If you plan quarterly goals as well, write them first with the prompts for writing OKRs; the weekly “max two goals” field should then point at a key result, which keeps each week connected to the quarter.
