Prompting for productivity

Where AI actually saves time

The biggest productivity gains from AI assistants usually aren't dramatic. They come from small, repeated tasks: turning a brain dump into a plan, pulling action items out of a meeting, sorting a pile of emails, or making a checklist you'd otherwise build from scratch.

The trick is to treat each of these as a reusable prompt. Write it once, refine it, save it, and reuse it every week.

Planning: from overwhelm to next steps

When a project feels too big, AI is good at breaking it down — as long as you tell it your real constraints. A generic plan assumes unlimited time and a typical situation. Yours isn't typical.

Notice the last instruction: "list anything that won't fit." Without it, the model may cram everything in and produce a plan that looks tidy but is impossible.

Meeting notes: decisions, actions, and open questions

Raw notes and transcripts are full of tangents. A structured prompt turns them into something people can act on.

Below are notes from a meeting. Turn them into a summary with these sections:

**Decisions made** — bullet list.
**Action items** — table with columns: Task | Owner | Due date.
**Open questions** — anything raised but not resolved.

Rules:
- Only include owners and due dates that are stated in the notes. If one is missing, write "unassigned" or "no date" — don't guess.
- Keep each bullet to one sentence.
- Don't include small talk.

Notes:
"""
[paste notes or transcript]
"""

The rule about not guessing owners matters. An AI summary that confidently assigns a task to the wrong person causes real problems. Always skim the summary against your notes before sending it to the team.

Email triage: sort first, then draft

When your inbox is overflowing, split the job into two steps: sort, then respond.

Here are [number] emails, separated by "---". For each one, give me:
1. A one-line summary.
2. A category: Urgent (needs reply today), This week, FYI only, or Can archive.
3. If it needs a reply, the key question I have to answer.

Then list the Urgent ones first.

---
[email 1]
---
[email 2]

Once you know what matters, draft replies one at a time, giving the AI your decision: "Reply saying yes to Thursday at 3pm, ask them to send the agenda beforehand, keep it under 60 words." You make the call; the AI handles the wording.

Checklists: never start from zero

Checklists are perfect for AI: they're structured, and a decent first draft is easy to improve. Give the context that makes it specific to you.

Create a checklist for [task or event].
Context: [who, where, when, any special conditions].
Group items by [phase / category / time before the event].
Mark the 3 items that are most commonly forgotten.
Format as a markdown checklist I can paste into my notes app.

For example, "a packing checklist for a 4-day work trip to a rainy city, carry-on only, with one client dinner" produces a far more useful list than "a packing list."

Make it a system

1. Save your best prompts. Keep a note with your planning, meeting, and triage templates.
2. Use the same format every time. Consistent output is easier to scan and paste into your tools.
3. Refine after each use. If the AI keeps missing something, add a line to the template.
4. Stay the decision-maker. The AI organizes and drafts; you decide priorities and commitments.