Friday review rituals collapse for two reasons. Either they take 45 minutes and you stop doing them, or they take 5 minutes but produce a recycled list that looks suspiciously like last week’s. The trick is to feed AI what you actually did, not what you planned to do, and to force a real “drop one” decision. Five minutes is enough if the inputs are honest.
TL;DR: Paste an honest braindump of what shipped and what stalled into any chat model, run the prompt below, and get back six tight sections including next week’s three priorities, one deliberate drop, and one honest flag. Use ChatGPT Plus, Claude Pro, or Google AI Pro (all $20/month or less as of June 2026) — model quality is not the bottleneck here, input honesty is.
The task
It is Friday afternoon. You want a short written reflection that names what you actually shipped this week, what stalled, and what your three priorities are for next week — including one priority you are deliberately dropping and one thing you need to raise honestly with someone (your manager, your co-founder, your future self).
When this is the right job for AI
- You have a usable record of what you did this week — calendar, commit log, a note app, anything tangible. AI cannot reconstruct your week.
- You are willing to write down at least one thing you stalled on. The reflection is useless if you only list wins.
- You will spend the 5 minutes. Even a great template fails if you let the week end without it.
- You want a structured output you can scan in 30 seconds on Monday morning.
What to feed the AI
- A raw braindump of this week: meetings that mattered, work that shipped, things that stalled, surprises
- Your originally planned priorities at Monday-start (so you can see what survived and what got swapped)
- One person or context you are accountable to (so the “honest flag” has a real audience)
- The general scope of your role for the next week (any planned absences, time-block constraints)
- One thing your gut already wants to drop, even if you have not said it out loud
Copy-ready prompt
You are turning my raw Friday braindump into a 5-minute weekly reflection.
This week, raw braindump:
- Monday planned priorities were: (1) ship onboarding v3, (2) finish the Q3 OKR draft, (3) interview 3 candidates for the senior eng role.
- What actually shipped: onboarding v3 (Tue), 2 of 3 candidate interviews (Wed and Thu).
- What stalled: OKR draft — got pulled into a production incident Wed afternoon, never reopened the doc.
- Surprises: the production incident took 4 hours and surfaced a monitoring gap. I drafted a follow-up but did not assign owners.
- Meetings that mattered: 1:1 with my manager Thu — they hinted Q3 scope is shrinking.
Accountable to: my manager (we have a 1:1 next Tuesday).
Next week scope: full week in office, no planned absences. Wednesday is heavy meeting day.
Gut wants to drop: writing the monitoring follow-up myself instead of handing it to the on-call rotation.
Write the reflection in six short sections:
1. **Shipped.** What actually got done. 2-4 bullets. Tense: past simple. No "made progress on."
2. **Stalled, and why.** The thing that did not ship plus the actual reason. One sentence per item. Be specific — "got pulled into incident" not "got busy."
3. **Next week, three priorities.** Numbered. Each priority names a verb (write, ship, interview, decide), not a topic. Tie one to the manager 1:1.
4. **Dropping deliberately.** One item. Not "deprioritized" — actually dropped, with the reason and what changes because of it.
5. **Honest flag.** One thing I need to surface to my manager next Tuesday that I would otherwise let drift. Write it as the sentence I would actually say.
6. **The 30-second Monday view.** A 2-line summary I can read at 9am Monday and act on without rereading the rest.
Rules:
- Do not invent items I did not include in the braindump.
- Do not let the "three priorities" repeat last week's failed list unchanged.
- The "drop" must be an actual drop, not a reschedule.Sample output structure
Shipped.
- Onboarding v3 went live Tuesday morning, monitoring dashboard updated.
- Completed 2 of 3 senior eng candidate interviews (the third candidate rescheduled, not a stall on me).
- Drafted the post-incident follow-up doc Thursday evening — needs owners assigned.
Stalled, and why. The Q3 OKR draft is still at headings only. The Wednesday production incident took 4 hours and I never reopened the doc afterward — context was lost and I deprioritized it past a fake “I will pick it up Friday” plan that did not happen.
Next week, three priorities.
- Write the Q3 OKR draft to a reviewable v1 by Wednesday EOD. This is the manager 1:1 input — the hint that scope is shrinking means the draft needs to land before the conversation moves without me.
- Assign owners to the monitoring follow-up by Tuesday 12:00. Two named owners and a check-in date, then off my plate.
- Run the third senior eng interview and write the hire/no-hire memo same day. Delay between interview and memo loses signal.
Dropping deliberately. I am not writing the monitoring follow-up myself. The on-call rotation owns the runbook — me writing it bypasses the team that has to live with it. Cost of dropping: the doc may take 3 extra days. Acceptable.
Honest flag for the Tuesday 1:1. “I want to talk about what shrinking Q3 scope means concretely — specifically whether the onboarding-v4 work I have been assuming I would lead is still on, because I am about to commit hiring decisions against that assumption.”
30-second Monday view. Three things this week: OKR draft v1 by Wed EOD, monitoring follow-up owners by Tue noon, third interview + hire memo same day. Raise the Q3-scope question with the manager Tuesday before it drifts.
How to refine
- Output reads like last week’s list → reject and require: “name what is actually different from last week’s three priorities; if nothing, call that out explicitly.”
- “Dropping deliberately” is a reschedule in disguise → strict rule: “the dropped item cannot appear elsewhere in this reflection, including the 30-second view.”
- Honest flag is too vague (“talk about workload”) → require: “write the actual sentence I would say, in quotes, including the specific topic.”
- Priorities are nouns not verbs (“the OKR draft”) → re-prompt: “each priority must start with a verb and name a Friday-checkable outcome.”
- 30-second view is longer than 2 lines → cap it; if AI keeps overflowing, prompt: “if you cannot say it in 2 lines, the priorities are not sharp enough.”
Which tool to use (June 2026)
A 5-minute reflection runs fine on a free tier — this is short, single-turn text. Where the paid tiers help is removing the braindump step so you are not retyping your week from memory.
| Setup | What it adds | Cost (as of June 2026) |
|---|---|---|
| Any free chat (ChatGPT Free / Claude Free / Gemini) | Paste braindump manually, run the prompt | $0 |
| ChatGPT Plus Project | Save the prompt as Project instructions; drop weekly notes in the same Project so context carries over | $20/mo |
| Claude Pro + Cowork | Connect Google Calendar, Gmail, and Slack; a scheduled task can assemble the braindump for you every Friday | $20/mo ($17 annual); Cowork is bundled |
| Google AI Pro | Gemini 3.1 Pro reads your Google Calendar and Gmail directly inside Workspace | $19.99/mo |
If your week already lives in Google Calendar and Slack, Claude Cowork is the lowest-friction path: type /schedule in Cowork to create a recurring Friday task that pulls the last seven days of calendar events and Slack threads, then runs this prompt automatically. Scheduled tasks are included on every paid Claude plan (Anthropic’s setup guide). The model you pick barely matters for a task this short — GPT-5.5, Claude Sonnet 4.6, and Gemini 3.1 Pro all handle it identically. What matters is the seven days of honest input.
Common mistakes
- Listing wins only. A reflection with no stall is a press release, not a reflection.
- Letting last week’s priorities roll over unchanged. If they did not move this week, something about the framing is wrong — change the verb, change the deadline, or drop them.
- Skipping the “drop.” Without a drop, the list grows every week and nothing actually exits.
- Writing the honest flag in third-person abstract. Write the actual sentence — if you cannot, you are not going to say it.
- Doing the reflection Monday morning. Monday morning is for executing the 30-second view; Friday afternoon is when the week is fresh.
FAQ
Is 5 minutes really enough? Yes, but only if you capture as you go. If Friday is the first time you try to remember the week, the bottleneck is not the 5 minutes — it is that you have nothing to feed the AI. Make a calendar scan plus a one-line “shipped today” note a daily habit; Friday is just synthesis.
What if I had a great week with no stalls? Reflect on a stall from earlier in the month. “No stalls” almost always means “I am not noticing them.” A reflection with no stall is a press release, not a reflection.
Why is “drop one” mandatory? The biggest failure mode of a Friday review is rolling last week’s list forward unchanged; do that for six weeks and you have forgotten why you are working on half of it. Forcing one deliberate drop makes you re-rank. If you genuinely cannot drop anything, demote one item to the week after next.
Can I do this on Sunday night instead of Friday? Sunday lets context decay and adds dread. Friday before logoff, while the week is still fresh, is meaningfully better. Monday morning is for executing the 30-second view, not writing it.
Won’t the AI write the “honest flag” too softly? Yes, by default. The prompt forces it to name the person, the specific topic, and the actual sentence you would say, in quotes. Drop any of those three and the model hedges. Reject vague output like “talk about workload.”
How long until this pays off? Around three weeks in you start seeing patterns — the same item stalling, the same drop you keep failing to make. Stack twelve of these and your quarterly review writes itself.
Related
- AI App Review Appeal Draft
- AI retention cohort analysis
- AI user interview question generator
- AI user segment targeting brief
- AI feature prioritization for indie apps
Tags: #AI writing #weekly-review #office-study #Productivity