Most people use AI tools the same way they use Google — a short query, a generic result, vague usefulness. The difference between saving 10 minutes a week and saving 3 hours comes down almost entirely to how you write the prompt. These 10 prompts are designed for real knowledge work, with examples of what each actually produces.
Short Answer: The AI prompts that save the most time are specific, contextual, and tell the AI exactly what format you want the output in. Vague prompts produce vague output you still have to rewrite. The 10 prompts below are built around the tasks that consume the most time for knowledge workers — email, summarisation, planning, research, and content repurposing — each with a real example of the output you can expect.
How to Use This List
Don’t copy these prompts verbatim. Each one is a template — adapt the bracketed sections to your actual situation, your actual tone, and your actual context. The principle behind every prompt here is the same: give the AI your role, the task, the format you want, and any constraints that matter. The more specific the input, the less editing the output needs.
All of these work in Claude (claude.ai) or ChatGPT (chat.openai.com) on free tiers. None require plugins or paid features.
Prompts 1–3: Email and Communication
Prompt 1: The Follow-Up Email
“Write a follow-up email to [name/role] who hasn’t responded to my [proposal/quote/request] in [X days]. Context: [one sentence about what the original message was about and what you need from them]. Tone: polite but direct. Length: 3–4 sentences. Include a soft deadline of [date] for their response.”
Why it works: Follow-up emails are hard to write because you’re managing two competing goals — not seeming pushy and not losing momentum. Giving the AI the context, tone, length, and deadline constraint produces a draft that’s almost always usable with minor tweaks.
Prompt 2: The Difficult Reply
“I need to reply to this email: [paste email]. The core issue is [your actual concern in one sentence]. I want to [achieve outcome] without [the thing you want to avoid — sounding defensive, burning the relationship, etc.]. Write three different versions: one direct and short, one warmer and more detailed, one that asks a clarifying question before committing.”
Why it works: Difficult replies stall because you’re not sure which approach is right. Getting three options forces a choice rather than a blank page. One of the three is almost always close to correct.
Prompt 3: Meeting Agenda from Notes
“Turn these rough notes into a clean meeting agenda: [paste notes]. The meeting is [length] minutes with [number] people. Goal of the meeting: [one sentence]. Format: numbered agenda items with time allocations. Include one ‘decision needed’ flag next to any item that requires a resolution in the meeting.”
Why it works: Most meeting agendas are either too vague (“discuss project”) or too detailed to be useful in the room. The time allocation constraint forces the AI to prioritise, and the decision flag makes it immediately clear where the meeting needs to produce an output.
Prompts 4–6: Summarising and Research
Prompt 4: Long Document Summary
“Summarise this document in three formats: (1) a one-sentence headline summary, (2) a five-bullet executive summary with the most important points, (3) a ‘what do I need to do’ section listing any action items or decisions implied by this document. [Paste document or key sections].”
Why it works: Asking for three formats in one prompt means you get the right level of detail for different contexts — the headline for a Slack message, the bullets for a briefing, the actions for your task list. It also forces the AI to read the document through three different lenses, which catches more than a single summary does.
Prompt 5: Research Synthesis
“I’m researching [topic] for [purpose — a decision, an article, a presentation, a meeting]. Here are [X sources / these paragraphs]. Synthesise the key points into a structured summary. Note where sources agree, where they contradict, and flag anything that seems like an outlier claim that needs verification.”
Why it works: Pasting multiple sources and asking for synthesis rather than summary produces something much closer to analysis. The contradiction flag is particularly useful — it catches the cases where one source is an outlier and you’d have missed it reading sequentially.
Prompt 6: Explain This Simply
“Explain [concept/document/technical term] to me as if I’m [the relevant non-expert: a smart person who doesn’t know this field / a new team member / a non-technical stakeholder]. Use an analogy if it helps. Keep it under 150 words. Then tell me the one thing I need to understand to not be confused in a conversation about this topic.”
Why it works: The “one thing to understand” constraint is the most valuable part. It forces the AI to identify what actually matters for someone who needs working knowledge rather than deep expertise, which is usually the actual goal.
Prompts 7–9: Planning and Thinking
Prompt 7: Weekly Plan from Task Dump
“Here’s my full task list for this week: [paste tasks]. My available working hours are approximately [X hours]. My top priority outcome for the week is [one thing]. Organise these into a realistic daily plan from Monday to Friday. Flag anything that probably won’t fit and suggest what to defer or delegate. Be honest about the capacity — don’t just schedule everything.”
Why it works: The “be honest about capacity” instruction is the key differentiator. Without it, AI tools tend to optimistically schedule everything, which produces a plan that breaks by Tuesday. Explicitly asking for a capacity-honest plan produces something more usable.
Prompt 8: Decision Framework
“I’m deciding whether to [decision]. The key factors I’m weighing are [list them]. My main concern is [one sentence]. Give me: (1) the strongest argument for doing it, (2) the strongest argument against, (3) what information would most change your recommendation, (4) your honest recommendation based on what I’ve told you.”
Why it works: AI is useful as a thinking partner precisely because it has no emotional stake in the decision. Asking for the strongest argument on both sides prevents you from getting a sycophantic answer that just agrees with the framing you used. The “what information would change your recommendation” part often surfaces the thing you haven’t actually thought through yet.
Prompt 9: First Draft of Anything
“Write a first draft of [document type: report, proposal, article, script]. Context: [2–3 sentences about the audience, purpose, and most important point to make]. Length: approximately [word count]. Tone: [professional/conversational/persuasive]. Structure: [any structure you want, or ask the AI to suggest one]. Include [any specific sections or elements you need].”
Why it works: This is the meta-prompt — the structure for getting a first draft of almost anything. The specificity of audience, purpose, tone, and structure is what makes the output usable rather than generic.
Prompt 10: The Meta-Prompt — Build Your Own
Once you understand the pattern behind effective prompts, you can build one for any recurring task. The formula is always the same:
[Your role or context] + [Specific task] + [Format and length of output] + [Tone or style constraints] + [Any specific elements to include or avoid]
The fastest way to improve a prompt that’s producing mediocre output: tell the AI what was wrong with the last response. “That was too formal — rewrite it in a more conversational tone.” “The summary was too long — cut it to five bullets.” “You missed the most important point which was X — redo it with that as the centrepiece.” Iteration on a bad output is almost always faster than starting again from scratch.
Common Problems and Fixes
The output is generic and feels like it could apply to anyone
You haven’t given enough specific context. Add: who the audience is, what they already know, what the specific situation is, and what you want them to do or feel after reading the output. Specificity is the most reliable lever for improving AI output quality.
The output is too long and needs heavy cutting
Add a word count constraint to the prompt: “under 200 words,” “no more than five bullet points,” “three sentences maximum.” AI tools default to comprehensive when length isn’t specified. A length constraint forces prioritisation, which usually improves the output beyond just making it shorter.
The AI keeps changing my intended meaning when I ask it to edit
Specify what you want changed and what you want preserved: “Edit this for clarity and concision. Do not change the key argument or the specific examples I’ve used. Only change wording that is unclear or unnecessarily wordy.” Without this instruction, AI editing tends to rewrite rather than refine.
Important: AI tools produce confident output regardless of accuracy. For any prompt involving facts, statistics, dates, or specific claims you plan to publish or send externally, verify the output against a source you trust. The prompts above are strongest for structuring, drafting, and synthesising — not for fact generation.
Before You Start Using These Prompts
Checklist:
- ✅ Save prompts that work well — you’ll reuse the same ones repeatedly
- ✅ Adapt the template to your specific context before running it
- ✅ Run the prompt manually once before adding it to any automation
- ✅ Fact-check output that includes statistics or specific claims
- ✅ Iterate on poor output rather than starting again — tell the AI what was wrong
- ✅ Build a personal prompt library — a simple doc with your best prompts by category
Related on Techtippr
AI Productivity Tools for Beginners: The Complete Guide
Frequently Asked Questions
Do AI prompts actually save meaningful time, or is it marginal?
For tasks you do repeatedly — follow-up emails, meeting agendas, document summaries, weekly plans — the time saving is material. A follow-up email that takes 12 minutes to draft from scratch takes 2 minutes with a good prompt and light editing. At five follow-up emails per week, that’s 50 minutes saved weekly from one prompt type alone. The compounding effect across several prompt categories adds up to several hours per week for most knowledge workers.
Which AI is best for productivity prompts?
Claude (claude.ai) produces the most natural, well-structured written output and is particularly strong for long-form drafting, summarisation, and nuanced communication tasks. ChatGPT is fast and strong on structured outputs (lists, plans, frameworks). For most of the prompts in this article, either works well — try both on the same prompt and use whichever output is closer to what you need.
Can I use these prompts for client work?
Yes, with one caveat: don’t paste confidential client information into a public AI tool if your client relationship or industry requires data confidentiality. For prompts that need to include sensitive context, either use a self-hosted AI model, use an enterprise plan with a data processing agreement, or abstract the sensitive details into generic placeholders before running the prompt.
How do I build a prompt library?
Start a simple document (Notion, Google Docs, or even a text file) with sections for each category: Communication, Summarisation, Planning, Research. Every time a prompt produces output you’re happy with, paste it into the library with a note on when to use it. After a month of active use, you’ll have 15–20 reliable prompts that cover most of your recurring tasks. Reference the library before starting a new task rather than writing a new prompt from scratch every time.
AI Tools for Beginners ↗
Automate Tasks with AI
Conclusion
The difference between AI tools feeling useful and AI tools feeling gimmicky is almost entirely in prompt quality. These 10 prompts are built around the tasks that consume the most time in a knowledge worker’s week — and each one is designed to produce output you can actually use with minimal editing.
The meta-skill is learning the pattern: role + task + format + constraints. Once that’s internalised, you can build a prompt for any recurring task in your work. Build the library, reference it habitually, and the time savings compound every week.