26 AI agent prompts for Personal Productivity

Copy-paste, beginner-friendly prompts to help you triage email, plan your week and summarize long documents. Each one is based on a real-world example and works with ChatGPT, Claude or an AI agent. 19 are free below, and 7 more are in AgentPlaybook Pro.

Free Personal Productivity prompts

My tasks: [paste]. My fixed commitments: [paste]. My energy is best in the [morning/afternoon]. Make a realistic weekly plan with time blocks, top 3 priorities, and what to drop or delegate.

Summarize this document in: a 3-sentence overview, 7 key points, any deadlines/numbers, and 3 questions I should ask. Document: [paste or upload]

Teach me [topic] as a complete beginner in 5 short lessons. Each lesson: simple explanation, real example, one exercise. Quiz me at the end and adapt to my mistakes.

Help me remove my personal and contact info from B2B contact-data sellers and people-search brokers (for example Apollo, ZoomInfo, RocketReach, Lusha, ContactOut, Seamless AI, and the big people-search sites). My details: name [name], emails [emails], phone [phone], city [city]. For each site: check whether I'm listed, find its official opt-out or data-deletion page, and either submit the request or give me the exact steps and link. Never create accounts, pay for anything, or share more info than the form needs without asking me first. Finish with a table: site, listed (yes/no/unknown), request sent (yes/no), date, and when to re-check.

Help our group plan [trip / dinner / tickets / carpool / event] in this chat. People: [names and any constraints]. Window: [dates or times]. Budget: [amount or 'flexible']. Preferences: [location, dietary, vibe]. Propose 2-3 options with pros/cons, then once we pick one, list who does what (book, pay, bring, drive) and a shared checklist. Ask before using anyone's personal calendar, contacts, or payment details, and don't book, message people outside this chat, or share one person's private info with the group unless they say yes.

Build me a simple pipeline in this folder that turns podcast episodes into a personal wiki. Input: a transcript (or audio file I'll transcribe) for each episode in [inbox folder]. For each episode: 1) write an episode page with the show, guest, date, and the 5 most useful takeaways; 2) pull out the main ideas and create or update one Markdown page per idea, linking back to every episode that mentions it; 3) add links between related ideas. Keep everything as plain Markdown files I can open in [Obsidian / any notes app]. Never delete or overwrite my own notes; only add to them. Show me the result for one episode first before running the rest.

Help me run a daily job search without scrolling job boards. My resume is at [path or paste]. My preferences: [roles, locations, remote/hybrid, seniority, must-haves, deal-breakers]. Using Pinloop (or equivalent job search CLI), pull new postings for [search terms] from the last [24 hours / today], judge each against my resume and preferences (no / weak / fair / strong) with a one-line reason, and add only 'strong' matches to a shortlist named [shortlist name]. Show me the shortlist as a table: company, title, score, reason, link. Don't apply anywhere, don't email recruiters, and don't change my profile. Wait for me to pick which roles to pursue.

Use computer use on my machine with approval before controlling each app. Goal: [e.g. submit this week's expense report / update slides / move data between two desktop apps]. Apps involved: [list]. Constraints: [e.g. never click Submit/Send/Pay without my OK; stay under $X; only use the sandbox account]. Step through the UI, narrate what you see, and pause for approval before irreversible actions. When done, report: screens visited, fields filled, confirmation numbers, and anything unexpected. If a dialog you don't recognize appears, stop and ask me.

Help me simplify my AI agent setup. Current setup: [list each agent or scheduled run, what it does, how often it runs, and how often I actually read or act on its output]. My active projects/clients: [list]. Do three things: (1) mark each agent or run as keep, merge, or turn off, with a one-line reason based on whether I actually use its output; (2) design one folder per project with a context file (goals, people, decisions, open tasks) and a running log, and tell me what goes in each; (3) write the instructions for a single assistant that sees all projects, keeps one master list of open tasks, and only speaks when I ask or when something is due. Keep the plan doable in one afternoon.

You're my proactive assistant. You can see [my email, calendar, Slack/Teams, and: list other apps]. Here's what matters to me: [paste: top priorities this month, key clients or people, deadlines, and things I never want to miss]. Each morning, send me a brief of no more than 5 items: what needs a decision from me today, meetings I should prep for, and anything overdue. During the day, check my apps every few hours and only message me if something is urgent, from a key person, or blocks a deadline; group everything else into the next morning's brief. For routine replies, draft them for me to approve. Before you send a message, make a payment, sign anything or delete anything, show me exactly what you're about to do and wait for my yes. Tell me how often you messaged me this week and ask if I want more or less.

You have access to a phone through [PhoneUse / your phone-control tool]. Task: [e.g. check my Instagram DMs from today and send Sam a 3-line summary on WhatsApp / compare the price of this grocery list in two delivery apps / book a ride from X to Y for 6pm]. Rules: Step 1: tell me which apps you'll open and what you'll do in each before you start. Step 2: only read, search and fill in; stop and show me a screenshot and wait for my yes before you send a message, book, buy, pay, post or delete anything. Step 3: never enter passwords or one-time codes yourself; if you hit a sign-in or code, stop and ask me. Step 4: when done, give me a short report of what you did, with the final screenshots, and anything you couldn't finish.

You're my household shopping assistant. Every [week / two weeks], build my order from [store]. My usual list: [paste staples, brands and quantities]. Household: [number of people, diet needs, allergies]. Rules: 1) Keep the total under $[limit]; if it would go over, drop the lowest-priority items first and tell me what you dropped. 2) Only swap a brand if the usual one is out of stock or at least [20]% cheaper, and list every swap. 3) Never change my payment method or delivery address. 4) Before checking out, show me the cart: item, quantity, price, and total, plus anything I seem to be running low on based on past orders. 5) Wait for my 'yes' before placing the order unless the total is under $[auto-approve amount].

You're my agent's coach. Below are transcripts or notes from past attempts at [type of task, e.g. writing client reports / fixing bugs in our app / preparing invoices], each marked WORKED or FAILED, plus any feedback I gave: [paste]. Write a lessons file under 400 words with: 1) my preferences (format, tone, length, what I always change); 2) rules and conventions that weren't written down but mattered; 3) gotchas in our tools, files or data that caused failures, and how to avoid them; 4) a short checklist to run before calling the task done. Only include lessons backed by the transcripts, and note which attempt each one came from. Skip one-off details. I'll paste this file at the start of every similar task, so keep it tidy and update it when new attempts contradict it.

Here is what I have from the last week: my sent and received emails [paste or connect], my task list [paste], and my meeting notes or transcripts [paste]. 1) List every open loop: things I promised to do, things others promised me, approvals waiting on me or on someone else, and questions nobody answered. For each, give who, what, the date it came up, and where you found it. 2) Mark each one as 'I can finish this now' (for example, a reply or a reminder), 'needs my decision', or 'waiting on someone'. 3) Draft the replies and reminders for the first group, ready for me to review and send. 4) For the 'needs my decision' group, write a one-line summary and the options. 5) Give me a short plan for the rest of the week in priority order. Don't send anything or change any records without my OK.

Help me set up a 'check the source, don't trust memory' rule for my AI assistant(s): [list assistants and what each does]. Keep all current permissions for sending, buying and publishing exactly as they are. 1) For each assistant, list the facts it relies on that can change: dates, prices, statuses, contact details, balances, order info. Next to each, name where the current value lives (for example [Google Calendar / Gmail / Notion / spreadsheet / store admin]) and say whether you can actually reach it. Flag anything missing or conflicting. 2) Review what each assistant remembers. Keep stable preferences and rules. For changing facts, propose replacing the value with where to check it, e.g. 'check Google Calendar for the next review date'. Show me the changes for one assistant first and wait for my OK. 3) Write this standing instruction for each: 'Before stating or acting on a fact that can change, check its source and say where and when you checked. If you can't reach it, say so, label the remembered value as old and unverified, and don't act on it.' 4) Pick three changing facts and test: did it check, what did it find, and did that differ from memory? Report results honestly; if you can't tell, say so.

Help me rebuild my timesheet for [date]. Here's what I have: Calendar: [paste that day's meetings and events]. Tasks: [paste tasks completed or updated, with times if shown]. Activity: [paste sent emails/messages subject lines with times, files edited, commits or anything else with timestamps]. My clients/projects: [list]. Do this: 1) Build a timeline of the day from start to finish in 15-minute blocks, and say which evidence each block is based on. 2) Turn it into proposed time entries: client, project, task description (client-friendly wording), start, end, duration, billable yes/no. 3) Mark any gap longer than 30 minutes as "unknown" and ask me what I was doing. Don't fill gaps by guessing. 4) Flag entries you are unsure about with a short reason. 5) Total billable and non-billable hours per client. Give me the entries as a table I can paste into [Toggl / Harvest / Clockify / my invoice]. I'll approve each one before anything is logged.

You are my task assistant. Here is my to-do list for today: [paste tasks, or connect your to-do app]. Part 1, rewrite: for each task, add three lines. Why: the reason it matters or the deadline. Done: exactly what finished looks like, so anyone could check it. Access: which apps, files, threads or accounts are needed. If you can't tell any of these from what I gave you, ask me one short question per task instead of guessing. Part 2, run: for each task that now has Why, Done and Access, do as much of the work as you can with the access I've given you, such as drafting the email, filling the invoice or preparing the calendar invite. Stop before anything that sends, pays, books or deletes, and leave it for me to approve. Finish with a list: done, ready for my approval (with what I need to click), and blocked (with what's missing).

I use these AI agents and tools: [list, e.g. an email agent, a coding agent, a social media scheduler, Zapier automations]. Below I've pasted everything from the last [24 hours] where they asked me something or are waiting on me: messages, notifications, approval requests, drafts awaiting review. [paste]. Make me one decision list. For each item: which agent, what it needs from me in one line, what's blocked until I answer, and a suggested answer I can approve with a yes. Sort by how much work it's blocking. Group anything I can answer the same way into one batch. Then suggest up to 3 standing rules I could give my agents so they stop asking me about the same kind of thing (for example, approve refunds under $20), and flag anything that should always stay with me. Don't send or approve anything yourself.

You are my office assistant. Golden rule: you NEVER delete, empty trash, or overwrite anything. If something should go, move it to a folder called 'Old' (create it if needed) and list what you moved. Also never send emails; only draft them. Every morning do this: 1) INBOX: sort new emails into [folders, e.g. Action today / Waiting on others / FYI / Newsletters], and give me a 5-line summary of what needs me today. 2) CALL NOTES: turn my rough notes in [folder or app] into clean notes with decisions, action items with owners and dates, and open questions. 3) FILING: save email attachments into the right folder using these rules: [e.g. invoices > Finance/2026/Month; contracts > Legal/Client name]. If unsure, put it in 'To file' and ask me. 4) FIRST DRAFTS: for anything in my task list marked [DRAFT], produce a first version in [Word / Excel / PowerPoint] using [template], marked DRAFT in the file name. 5) Finish with a short log: what you sorted, filed, moved to Old and drafted, and anything you weren't sure about.

More in AgentPlaybook Pro

• Inbox triage in draft modeSort email and draft replies without sending anything automatically.
• AI executive assistant setupSet up an agent that handles scheduling, reminders, and follow-ups.
• Have an agent chase a company over a faulty productHand the back-and-forth of a repair, warranty, or refund claim to an agent that writes to the company's support, answers their follow-up questions, and checks with you before sharing personal details, like Yantra's Lucy assistant getting a broken Dyson Airwrap fixed from a single text.
• Nightly cleanup of the notes your agents shareUse one notes vault as shared memory for all your AI agents and have an agent tidy it every night, fixing out-of-date facts and linking loose notes, like Eugeniu Ghelbur's Obsidian setup shared by eight agents including Claude Code and Codex.
• Get coached in real time during a customer service callBuild a small helper that listens to a call you've told the other side is being recorded, like a health insurance dispute, and shows you what to say next in a second or two. Kevin Labianco had Claude build one for his own insurance call and read its suggestions out loud as he went.
• Coordinate a family member's care across doctors, bills and follow-upsUse an agent as a healthcare advocate for a parent, partner or child: it keeps track of several specialists, catches missed follow-ups and test results, handles prescription refills and renewals, and helps challenge surprise bills and insurance denials. Vene launched on Oct 21, 2026 as an 'AI advocate for healthcare'; its founder says over 5,000 families use it, including his mom to coordinate her primary care doctor and endocrinologist and renew prescriptions.
• Plan your day around one morning question and review it at night with Benjamin Franklin's two-question scheduleBenjamin Franklin printed his daily schedule in his autobiography: at 5 a.m. he asked himself what good he would do that day, and at night he asked what good he had done. You can turn that into a simple routine with an AI assistant, planning the day around one meaningful goal each morning and doing a short honest review each evening. On Oct 22, 2026 the X account AI In Public shared a prompt that turns Grok Bot into a daily planner built on Franklin's two-question schedule.

See Pro pricing →

Guides that include Personal Productivity prompts

Other categories