TL;DR
Paste your full, messy task list into a reasoning model (GPT-5.5 Thinking, Claude Sonnet 4.6, or Gemini 3.1 Pro all work), give it your real hours and your office constraints, and force a hard distribution: at most 3 do-today, at least 5 defer-with-date, the rest drop or delegate. Use the copy-ready prompt below. The model is good at applying a framework consistently and at the “drop” verdict humans avoid; it is blind to office politics, so feed those in. If you want the result auto-blocked onto a calendar, hand the ranked list to a scheduler like Motion or Reclaim.ai. Run it daily during sprint weeks, weekly otherwise.
The task
It’s 8:47 AM. Inbox is 47 unread. Linear shows 18 open tickets assigned to you. Slack has 6 unread channels and your boss DM’d at 7 AM. Your calendar says 4 hours of focus blocks between meetings, and your honest assessment is you have 7 hours of real work today before the day collapses. Your gut response is to start with whatever is loudest, which is how you got here. You want a stack-ranked list of what to actually do, what to defer with a date, and what to drop entirely, plus a specific plan for the first 60-minute block while your attention is still good.
Which AI to use for this
You do not need a dedicated app to stack-rank a list. Any current reasoning model does it well, because the job is judgment over text, not scheduling. As of June 2026:
| Tool | What it costs | Best for | Note |
|---|---|---|---|
| ChatGPT (GPT-5.5 Thinking) | Free tier (tight limits); Plus $20/mo | Fast, sharp distribution; pick the “Thinking” mode for ranking | Free US tier shows ads since Feb 2026 |
| Claude (Sonnet 4.6) | Free tier; Pro $20/mo | Long, messy paste-ins; 1M-token context handles a week of Slack | Pro bundles Claude Code + Cowork |
| Gemini 3.1 Pro | Google AI Pro $19.99/mo | If your tasks already live in Workspace (Gmail, Docs, Tasks) | 1M context; was “Gemini Advanced” before early 2026 |
| Motion | Pro AI from $29/mo (individual) | Auto-blocking the ranked list onto your calendar | Calendar-first; its algorithm re-plans your day automatically |
| Reclaim.ai | Free tier; paid from $10/seat/mo | Auto-scheduling tasks pulled from Todoist, Asana, ClickUp, Jira, Linear | Swap a task’s priority and it re-arranges your calendar |
The split that matters: use a chat model to decide the ranking (this article’s job), then optionally use Motion or Reclaim.ai to place the do-today items on your calendar. Schedulers are weak at the “should this exist at all” call; that is exactly where a chat model with your office context wins.
For most people, paste into GPT-5.5 Thinking or Claude Sonnet 4.6 and skip the dedicated app entirely.
Where AI helps, and where it does not
AI is good at applying a prioritization framework (impact × urgency, the Eisenhower matrix, or RICE) consistently across a messy list, forcing a distribution instead of letting everything be “high priority,” and surfacing the dependencies you forgot (the PR that unblocks two reviewers, the email that unblocks the legal call). It also handles the “drop” verdict better than most humans. Humans hesitate to drop; AI does it cleanly when told to.
What AI cannot do: know your office politics. It does not know that your VP just got nervous about the launch, that your teammate is blocked on you specifically (not just “someone”), or that your skip-level is reading this week’s update. Feed those signals in as constraints. AI also cannot judge whether a task is “actually urgent” or “someone else’s urgent pretending to be yours”; challenge it on every urgency claim.
A specific failure mode: AI tends to mark too many things “do today” when the list is long. Tell it explicitly: force a distribution, at most 3 do-today, at least 5 defer, the rest drop or delegate. If do-today has more than 3, it is not prioritizing.
What to feed the AI
- Full task list, pasted from anywhere (Linear, Asana, Notion, inbox subject lines, Slack threads; messy formatting is fine)
- Real hours available today after meetings (be honest, “7 hours” is rarely true)
- Hard constraints: things you cannot drop (committed deadlines, people waiting on you specifically, calendar holds you cannot move)
- Office context: what your manager will notice if you skip, what your team is blocked on
- The one task you have been avoiding (every list has one); name it so the model can decide whether avoidance is the right call
- Your energy level honestly. High-energy 60-min blocks should go to cognitive work; low-energy time goes to admin
- The single “if today fails, why” item: what failure would make tomorrow much worse
- A separate “personal” list if you are combining work and life; the framework still works
Copy-ready prompt
Replace each [bracket] with your own text before sending.
Stack-rank my task list for the next [hours] hours.
Full list (paste, messy is fine): [paste]
Hard constraints (cannot drop): [paste]
Office context, who is blocked on me, what my manager will notice: [paste]
The task I have been avoiding: [paste]
The "if today fails, why" item: [paste]
My energy distribution: [paste, e.g. "good 9-11, ok 1-3, drained 3-5"]
Framework: impact x urgency.
For each task, return:
- Rank (1, 2, 3...).
- Verdict: do-today / defer-with-date / drop / delegate-to-whom.
- One-line rationale citing impact AND urgency separately ("blocks 2 reviewers + needed for end-of-day calibration").
- A flag if this task unblocks others ([UNBLOCKS: list]).
- An estimate in minutes (be realistic, most tasks take longer than expected).
Rules:
- Force a distribution: at most 3 do-today, at least 5 defer-with-date, the rest drop or delegate. More than 3 do-today means you are not prioritizing.
- Every defer needs a specific date. "Later" is not a date.
- Drop is a real verdict. Use it. About 30% of inbox items deserve drop.
- For do-today items, sequence by both dependency (unblockers first) and energy (cognitive work in the high-energy hour).
End with a 60-minute morning-block plan:
- The single task to start with at [time].
- Why this task, not the noisy one.
- What success looks like at the 60-min mark.
- What to do if the task is bigger than 60 min (stop and re-plan, do not push through).Shorter variant, daily 5-min standup
Yesterday I did: [list]. Today I have [hours] hours. Inbox highlights: [paste]. Pick the 3 do-today items. Defer everything else with a specific date. No "later." Lead with the unblocker.Sample output
A useful stack-rank row: “1. Ship the migration PR, do today, ~90 min, blocks 2 reviewers and required for end-of-day calibration meeting. [UNBLOCKS: design QA, data validation].”
A useful defer-with-date entry: “Refactor old config file, defer to Jun 19 (Friday focus block), high effort (~3 hours), no current blocker, but worth doing before Q3 planning. Calendar-block it.”
A useful drop entry: “Reply to recruiter email, drop, you have a job you like and this recruiter has emailed 3x in 6 months without a real role. Mark as read, do not reply. If they email a 4th time with a specific JD, reconsider.”
A useful 60-min morning block: “9:00-10:00, open the migration PR file in your editor and finish the diff. Do not open Slack first; the unblocking value is higher than anything you will see in Slack. Success at 10 AM: PR is committed and review-requested. If the diff is bigger than expected, stop at 10 AM, post a ‘I’ll have this by 2 PM’ update in the channel, and come back after lunch, do not push through and lose the next 2 hours.”
A useful avoidance call: “The task you have been avoiding is the contract review with Legal. Verdict: do-today at 10 AM. Avoidance has cost you 4 days already, and Legal closes at 5 PM Fridays; if you push it again, the deal slips a week. 30 minutes.”
How to refine the output
- Force the distribution: “Re-check: at most 3 do-today, at least 5 defer-with-date. If do-today has 4+ items, the 4th becomes a defer with a date. The constraint matters more than my objection that everything is urgent.”
- Defer needs a date: “Re-read defer entries. Every defer must have a specific date. ‘Next week’ is not a date; ‘Jun 19’ is. If the date is not specific, the item goes back to drop or do-today.”
- Use drop liberally: “Re-read the drop section. About 30% of inbox tasks should drop; if you have under 20%, you are too cautious. Drop is a verdict, not a failure. Common drop candidates: 6+ month-old emails, FYI Slack threads, recruiter outreach without a specific JD.”
- Surface dependencies: “Re-tag every task with [UNBLOCKS: list] if it unblocks others. Tasks that unblock others jump to the top of do-today, regardless of personal interest.”
- Anchor the morning block: “Re-check the 60-min plan. The first task must be high-impact and cognitive, not admin. If it is email or meeting prep, move it to a later block; morning energy is your scarcest resource.”
Common mistakes
- Padding the do-today list with easy items to feel productive. Finishing 5 easy tasks beats finishing 0, but finishing 1 hard task beats finishing 5 easy ones; force the distribution
- Defer as a black hole. Without a specific re-look date, deferred tasks live forever in your guilt pile; every defer gets a calendar entry
- Letting the model treat all “urgent” tasks as equal. Challenge each one; “my boss said this is urgent” is data, “I feel like this is urgent” usually is not
- No dependency flags. The PR that unblocks two reviewers is worth more than 5 self-contained tasks; tag unblockers explicitly
- Skipping the avoidance task. Every list has one; the model should call it out and decide whether avoidance is the right call (sometimes it is, when the task is genuinely lower priority, but usually it is not)
- Putting cognitive work in low-energy slots. The 3 PM crash is not when you should write a strategy doc; do admin then, write in the morning
- Running this once a week and never updating. The list goes stale; daily during sprint weeks, weekly otherwise, with a 5-min re-rank at any major shift (new urgent ask, day blown up)
- Treating it as an assistant who decides. The AI proposes; you commit. The decision still requires your office context and your honest energy assessment
FAQ
- Which model should I actually use?: Any current reasoning model handles this. As of June 2026, GPT-5.5 in “Thinking” mode and Claude Sonnet 4.6 are both strong; pick whichever you already pay for. Claude’s 1M-token context is handy if you paste a whole week of Slack and tickets at once. If your tasks live in Gmail and Google Tasks, Gemini 3.1 Pro can read them in place.
- Do I need Motion or Reclaim.ai for this?: No. Those are calendar-first schedulers (Motion Pro AI from $29/mo, Reclaim.ai free tier plus paid from $10/seat/mo as of June 2026). They place tasks on a calendar but are weak at the “should this exist at all” judgment. Use a chat model to decide the ranking, then optionally hand the do-today items to a scheduler to auto-block.
- Does this work for personal tasks too?: Yes. The framework holds: swap “blocks 2 reviewers” for “kid pickup,” “manager will notice” for “partner has been asking.” Combine work and life lists if your day blends them; force the distribution across the combined set.
- How often should I run this?: Daily during sprint weeks or any week where your list is shifting hourly. Weekly otherwise. Less than weekly and the list goes stale: defer dates lapse, drop candidates pile up, and the framework loses its bite.
- What if my manager keeps adding new “urgent” items?: Two moves. First, when they add an item, ask “what should this displace?” The answer reveals true urgency. Second, run the prioritization in front of them once; seeing the 3 do-today + 5 defer + rest dropped is what builds shared trust on capacity.
- The model is too conservative on drop. What changes?: Add: “About 30% of an inbox-style list deserves drop. If your drop count is under 25% of the list, you are too cautious. Drop is a verdict, not a failure mode. Common drop candidates: 6+ month-old emails, FYI Slack threads, vague ‘when you have time’ asks.”
Related
- AI weekly planning
- AI feature prioritization
- Workload Prioritization Prompts: 12 Templates to Decide What Not to Do
- AI Drafts Your Project Status Update
- AI Habit Tracker
Tags: #AI writing #Office #Workflow #Task priority #Planning