Specific prompt text, not a smarter AI, is what turns occasional use into a reliable shortcut. Every Tuesday for my first year freelancing, the newsletter draft ate 120 minutes.
My five bullet points were ready by noon, but the jump to subject line, preview text and three body sections was the work that took the afternoon.
One structured prompt in ChatGPT returned a complete draft in 30 seconds. Fifteen minutes of edits later it went out. The same output came back every week after that, and the draft stopped being the reason the newsletter ran late.
Over four months I timed every recurring task that ate my afternoons as a content strategist handling writing, project coordination and social media for three small-business clients.
Each one got a prompt, refined until the output needed only light edits before use. Every prompt here uses [bracketed slots] for your own material.
Claude Pro and ChatGPT Plus both start at $20 per month in the US as of October 2026. Check each pricing page before paying, because free-tier limits change often. Find the row where 'By Hand' hurts most and run that one prompt today.
| Task | By Hand (min) | AI Output Time | Review & Edits (min) | Best Tool |
|---|---|---|---|---|
| Weekly Email Newsletter Draft | 120 | 30 sec | 15 | ChatGPT |
| Cold Outreach Email (3 versions) | 45 | 20 sec | 10 | Claude |
| Performance Review Paragraph | 60 | 25 sec | 15 | ChatGPT |
| Job Description Under 400 Words | 45 | 20 sec | 10 | Claude |
| Timed Meeting Agenda | 30 | 10 sec | 5 | ChatGPT |
| Difficult or Rejection Email | 30 | 10 sec | 5 | Claude |
| LinkedIn Post from Bullet Points | 30 | 15 sec | 8 | ChatGPT |
| Blog Post Outline with SEO Signals | 60 | 30 sec | 15 | ChatGPT |
| 30-Day Social Media Calendar | 90 | 45 sec | 20 | ChatGPT |
| Product Description with Benefit Bullets | 45 | 20 sec | 10 | Claude |
| Long-Form Content Repurposed to Social Posts | 60 | 30 sec | 15 | Claude |
| Content Brief for a Freelance Writer | 60 | 30 sec | 15 | ChatGPT |
| FAQ Section from Any Document | 45 | 60 sec | 10 | Claude |
| PDF or Long Report Summary | 90 | 90 sec | 15 | Claude |
| Competitor Research Digest and Gap Table | 120 | 45 sec | 20 | Gemini |
| Meeting Notes to Action Items | 40 | 20 sec | 8 | ChatGPT |
| Sanitize-Before-You-Paste Privacy Prompt | 30 | 30 sec | 10 | Claude |
| One-Page Project Brief | 60 | 30 sec | 12 | ChatGPT |
| Case Study Draft from Project Notes | 120 | 60 sec | 25 | Claude |
| Weekly Planning and Daily Priority | 30 | 30 sec | 10 | ChatGPT |
1. Weekly Email Newsletter Draft Prompt
One setup session of about 20 minutes fixes the newsletter for every week after. The first time I ran this prompt, I fed ChatGPT a real Tuesday: five bullets about my client's product news, conversational voice, audience of small retail owners.
Back came a subject line, preview text, a 150-word lead story, two 60-word items and a 30-word call to action. Thirty seconds of AI output, then 15 minutes of my edits. The 120-minute Tuesday was gone.
Prompt to copy
You are an email newsletter writer for [industry or niche]. I will give you: (1) three to five bullet points of news or updates from this week; (2) my brand voice: [formal / conversational / friendly]; (3) my audience: [describe them in one sentence]. Your task: 1. Write a subject line and preview text of 40-50 words. 2. Write the newsletter body in three short sections: a lead story (150 words), two brief items (60 words each), and a closing call to action (30 words). 3. Flag any claim I should fact-check before sending. Output format: subject line / preview text / body sections labeled clearly. Rules: no filler phrases like \"In today's fast-paced world\"; match the voice I gave you exactly. Here are my bullet points: [paste bullet points].
The draft sounds confident even when it read my voice wrong. Once it opened warmly and chatty for a client whose past newsletters are dry and direct. Now I read the first output beside a past issue I liked before it reaches subscribers.
💡 Tip: Save the filled-in prompt as a template in your notes app. Next Tuesday, update only the bullet points.
2. Cold Outreach Email in Three Versions
The tedious part of cold email is producing one version, sending it, and never knowing whether a different angle would have landed. My first outreach email to a new prospect took 45 minutes for a single angle.
In Claude it comes back in 20 seconds. Three versions, each with a subject line and one low-friction ask: short and direct under 80 words, problem-led under 120 and social-proof-led under 120. Ten minutes of edits, and I have three testable angles in 30 minutes.
Prompt to copy
Act as an experienced B2B sales writer. I need a cold outreach email to [job title] at [type of company]. Context: [one sentence about what I do and what problem I solve]. Their likely pain point: [describe it]. My ask: [what I want them to do – one specific action]. Write three versions: (A) short and direct, under 80 words; (B) problem-led, under 120 words; (C) social-proof-led, under 120 words, using this proof point: [paste a result or testimonial]. Each version needs a subject line. Do not use \"I hope this finds you well\" or any similar opener. End each version with a single low-friction call to action.
3. Performance Review Paragraph from Notes
A performance review is an internal record, not a message to a warm audience or a stranger's inbox. The words stay in the file. They have to satisfy HR and land fairly with the person being reviewed.
My old way: 60 minutes turning rough bullet-point observations into polished, specific language, then a second pass to make sure nothing read as vague praise. HR reviewers flag generic summaries, and they are right to.
The new way: I paste those same observations into ChatGPT and hold the notes to concrete examples. Twenty-five seconds later I have a 100-130 word summary paragraph, a two-sentence rating justification and three suggested goals for next year.
Prompt to copy
You are an HR writing coach. I will give you rough notes about an employee's year and you will turn them into a polished performance review. My notes: [paste your bullet-point observations]. Their role: [job title]. Review tone required by my company: [formal / balanced / developmental]. Write: (1) a summary paragraph of 100-130 words covering achievements, impact and one growth area; (2) a suggested rating justification of two sentences; (3) a list of three suggested goals for next year. Rules: use specific language, not vague praise; do not invent facts not in my notes; flag with [CHECK] any claim that needs my verification before submitting.
Vague notes produce generic praise. The moment I supplied a real example in the notes slot, the output named what the employee actually did instead of calling them a team player.
Fifteen minutes of edits on the [CHECK] flags, and the task lands under 45 minutes.
4. Job Description Under 400 Words
A complete job posting came back to me in 20 seconds, structured and under 400 words. The requirements list I fed Claude was messy, a stack of must-haves typed in no order after a departing hire left the role open.
Prompt to copy
Act as a talent acquisition specialist. Write a job description for a [job title] role at a [company size and industry] company. Requirements I have so far: [paste your list]. Compensation: [range or competitive]. Location: [remote / hybrid / on-site, city]. Format: (1) two-sentence role summary; (2) what you will do – five to eight bullet points; (3) what you will bring – five to seven bullet points split into required and nice to have; (4) one short paragraph on culture or mission. Keep it under 400 words. Flag any requirement that might screen out qualified candidates unnecessarily.
The prompt names the job title, company size and industry, then the requirements, pay range and location. It asks for a two-sentence summary, five to eight duties, and requirements split into required and nice to have.
Where it fails: the AI copies selective or gender-coded phrasing straight from your list. The prompt flags requirements that might screen out good candidates, but read every flag before posting.
My 45 minutes by hand covered sorting tiers and trimming word count. Now it's 20 seconds plus 10 minutes of review, under 35 total. A performance review describes the person already in the role; this post is what brings the next one in.
💡 Tip: Read the bias flags first — they are the most legally important part of the output.
5. Timed Meeting Agenda with Pre-Read Note
Watch the time estimates. ChatGPT fills every slot optimistically, and a 30-minute agenda built on guesswork still overruns. Paste the talking points, meeting length, roles and goal, then adjust each allocation against how your team actually argues.
Prompt to copy
You are a meeting facilitator. Turn these talking points into a timed meeting agenda. Talking points: [paste your list]. Meeting length: [X minutes]. Attendees: [list roles, not names]. Goal of the meeting: [one sentence]. Format: a table with columns – agenda item, owner (by role), time allocation in minutes, and desired outcome for that item. Add a two-sentence pre-read note I can paste at the top of the calendar invite. Flag any item that looks like it belongs in a separate meeting rather than this one.
What you get back: A table with each item, its owner by role, minutes and the outcome you want. Below it sits a two-sentence pre-read for the calendar invite, plus a flag on any topic that deserves its own meeting.
Where it earns its place: One flag pulled a budget debate out of a 30-minute session before anyone sat down. Ten seconds of output, five minutes of adjusting, under 20 minutes total.
6. Difficult or Rejection Email Prompt
Thirty minutes is what one rejection email cost me, because every draft read either too cold or too evasive. Reusing a vendor, turning down a candidate, ending a contract: the wording stakes are high enough that most people wait until it is urgent.
Where the prompt stops helping: the [situation] slot. Type three vague sentences and you get a vague email back. The one honest reason has to be specific in the input, even when typing it feels uncomfortable. Fill it plainly and Claude drafts in 10 seconds.
Prompt to copy
Act as a professional business writer skilled at delivering unwelcome news with clarity and respect. Situation: [describe it in one to three sentences – for example: declining a vendor, ending a contract, rejecting a candidate]. My relationship with this person: [client / vendor / applicant / colleague]. Tone needed: [direct / warm / formal]. Write the email in under 150 words. It must: (1) state the decision clearly in the first sentence; (2) give one honest, non-apologetic reason; (3) close with a specific and genuine next step or good-faith gesture, not a hollow phrase like best of luck. Do not use the phrase at this time.
When to skip the tool entirely: if the person will read the email as a personal betrayal, or the decision still needs a phone call first, write it yourself.
For the routine no, the prompt caps the email at 150 words, opens on the decision and closes with a real next step, so total time falls under 20 minutes.
Prompt That Works vs. Prompt That Wastes Time
Prompt That Works
- Names the tool's role: 'Act as a B2B sales writer'
- Fills every bracketed slot with real material before sending
- Sets a hard word limit on each output section
- Includes an explicit flag instruction for claims to verify
- Bans the specific filler phrases the AI defaults to
Prompt That Wastes Time
- Opens with 'Write me a…' and nothing else
- Leaves bracketed slots blank or uses placeholder text
- Sets no length constraint — output sprawls
- No flag instruction — hallucinations go unnoticed
- Vague tone direction — AI defaults to corporate filler
7. LinkedIn Post from Bullet Points
Prompt to copy
Act as a LinkedIn ghostwriter who writes in a clear, direct, non-corporate voice. I will give you bullet points from something I learned, did or observed recently. My bullet points: [paste them]. My profession: [one line]. Audience: [who follows me – roles or industries]. Format: opening hook of one to two lines that does not start with 'I'; three to five short paragraphs with one idea each; a closing question or call to action. Length: 150-220 words. Do not use hashtags in the body. Suggest three hashtags to add at the end. Flag anything that sounds like a humblebrag or a cliche.
That is a different risk from a private difficult email. A rejection goes to one person. A post goes to hundreds, and the failure mode is identical: it sounds like everyone else's.
This prompt turns a few raw bullet points into a post with a hook that avoids the word "I," single-idea paragraphs, a closing call to action and three hashtag suggestions. ChatGPT returned mine in 15 seconds.
By hand, the same post cost 30 minutes. Most of that went to finding a hook and dragging the length into the 150-220 word range. The catch is motivational framing.
The model still slips into "lessons learned" structure and humblebrag phrasing. The prompt bans the opening "I" and asks it to flag the brags, but read that hook line yourself before posting. Review took me 8 minutes.
That flag has caught at least one post of mine that would have landed badly. Budget 8 minutes of review, and the total task comes in under 25 minutes instead of 30.
8. Blog Post Outline with SEO Signals
Planning a blog post that can rank starts at the outline, not the keyword tool. Mine used to take 60 minutes: five title options, the H2 and H3 structure, a word count per section, internal link targets, and one data point worth sourcing.
Prompt to copy
You are a content strategist and SEO writer. Create a detailed blog post outline on the topic: [topic]. Target reader: [describe them in one sentence]. Primary keyword: [keyword]. Search intent: [informational / comparison / how-to]. Format: (1) five title options with the keyword near the front; (2) a full outline with H2s and H3s, each with a one-sentence description of what that section will argue or show; (3) a suggested word count per section; (4) three internal-link opportunities I should look for on my site; (5) one data point or study I should find to support the main claim – mark it [WRITER TO SOURCE]. Flag any section where the argument is thin without a real example.
I fed the topic, reader, keyword and search intent into ChatGPT and had a full outline back in 30 seconds. The titles came with the keyword near the front. Every thin section carried a flag, and the data point came marked [WRITER TO SOURCE].
That marker is the catch: the AI never delivers a verified figure. I replace each one with a real number before drafting, which reads the same as the 15 minutes I spend on the rest.
The cost: 60 minutes by hand, 30 seconds of AI output, 15 minutes of my checks. Under 50 minutes total, with no separate keyword session.
9. 30-Day Social Media Calendar Table
Prompt to copy
Act as a social media strategist. Create a 30-day content calendar for [platform: LinkedIn / Instagram / X / TikTok]. My business or brand: [describe in two sentences]. Content pillars I want to use: [list two to four themes]. Posting frequency: [X times per week]. Format: a table with columns – date, content pillar, post type (tip / story / poll / carousel / video idea), and a one-sentence content summary. Include at least two engagement posts (polls or questions) and one repurposed-content slot per week. Do not write the full post copy; give me the calendar plan only.
One outline covers one post. The same structured logic covers thirty in a single run. This prompt returns a full table with a date, a pillar, a post type and a one-sentence summary for every slot.
Engagement posts and repurposing slots come built in by its own rules. ChatGPT produced my month in 45 seconds.
Last November I did this by hand over one long afternoon. Choosing pillars, assigning post types and writing 30 summaries cost 90 minutes. Three of the planned posts stopped making sense by mid-month, and the gaps showed.
The catch is that the AI fills every day with something plausible. It does not know whether you actually have a video to shoot or a photo to pull.
Check each post type against what you can really produce before you commit the month. That review took me 20 minutes.
Two times: 90 minutes by hand against 45 seconds of AI output worse, plus 20 minutes of your checking, which lands the whole month under 70 minutes.
⚠️ COMMON MISTAKE
Never Paste Real Client Files Without Sanitizing First
On free and personal plans, chats may be used to improve the company's models unless you switch that off in account settings. Run the sanitize prompt in item 17 before pasting any document that contains client names, account numbers or anything marked confidential.
10. Product Description with Benefit Bullets
Forty-five minutes for one product description, plus a second pass to strip out the 'best-in-class' and 'clinically proven' lines that crept in during the draft.
The prompt builds the same job in 20 seconds in Claude: an 8-12 word headline, a benefit lead, feature bullets and a call to action.
Prompt to copy
You are a direct-response copywriter. Write a product description for [product name]. What it does: [one sentence]. Key features: [paste your list]. Target buyer: [describe them]. Where it will appear: [website product page / Amazon / print catalog]. Tone: [premium / friendly / technical]. Write: (1) a headline of eight to twelve words; (2) a lead paragraph of 40-60 words hitting the top benefit first; (3) four to six feature-benefit bullet points in the format Feature — what it means for you; (4) a closing sentence with a call to action. Flag any claim — such as best, fastest or clinically proven — I would need to substantiate before publishing.
The catch: Claude writes a superlative when the tone slot is vague, so the prompt makes it flag every claim you'd have to prove. The flag list helps, but you are the one legally responsible for what goes live.
By hand: 45 min. AI output: 20 sec. Review and edits: 10 min. Total: under 35 minutes.
11. Long-Form Content Repurposed to Social Posts
Paste it into Claude as long-form content, not notes to write from.
Pulling those five standalone posts took 60 minutes by hand: rereading the piece, picking angles, rewriting each for the platform.
The prompt asks for the five strongest ideas, turned into self-contained posts, each with a one-line note on why it stands alone and a visual suggestion.
Prompt to copy
Act as a social media repurposing editor. I will paste a long-form piece of content below. Your job: 1. Extract the five strongest standalone ideas. 2. Turn each into a self-contained post for [platform]. Each post: [character or word limit for that platform]. Format for each: the post text, a one-line note on why this idea works as a standalone, and one suggested image or visual description. Do not just quote the original – rewrite for the platform voice and format. Here is the content: [paste article, transcript or notes].
The catch: it picks the five most quotable lines, not the five most useful ideas. Read each post as someone who never saw the original, and if it needs the article to land, rewrite it yourself.
The answer comes back in 30 seconds, review takes 15 minutes, and a finished article yields five posts in under 50 minutes, with no new ideas required. That makes it the setup worth building once.
12. Content Brief for a Freelance Writer
Commissioning content means handing a writer the title, the keyword, the section outline, competitor links, tone and word count.
Doing all that by hand took me 60 minutes, and the writer still emailed questions the next day. Those questions added half a day across three clients, every time.
The tedious part is that most of the 60 minutes went to assembling a document, not thinking. So I built the brief prompt below.
It returns a goal, an H2 outline, five data points flagged [WRITER TO SOURCE], three competitor URLs, tone notes and a word count. ChatGPT wrote it in 30 seconds; I reviewed it in 15.
Prompt to copy
You are a content director briefing a freelance writer. Write a content brief for an article titled: [working title]. Target keyword: [keyword]. Search intent: [informational / comparison / how-to]. Reader: [one sentence]. Angle that makes this different from the top results: [your one-sentence angle]. The brief must include: (1) goal of the piece in one sentence; (2) outline with H2s and a one-line description of what each section argues or shows; (3) five facts or data points the writer must include – mark each [WRITER TO SOURCE]; (4) three competitor URLs to beat; (5) tone and voice notes; (6) two or three internal links to suggest; (7) target word count. Flag any section where the angle is thin without a real example or data point.
The catch is the [WRITER TO SOURCE] flags. Every one is a fact ChatGPT could not verify, and deleting them to tidy the brief leaves research gaps a writer will paper over.
Keep them in. By hand: 60 min. AI output: 30 sec. Review: 15 min, mostly replacing the flags.
Content Brief — At a Glance
⏱️ By hand
60 min
🤖 AI output time
30 sec
✏️ Review & edits
15 min
🔗 Competitor URLs included
3
🚩 Data placeholders flagged
5
🛠️ Best tool
ChatGPT
Net time saved
A brief that used to generate a full day of follow-up questions now produces near-zero back-and-forth when the writer opens it on day one.
13. FAQ Section Generated from Any Document
Twelve plain-English questions came back ordered by importance, each answered in two to four sentences. A policy PDF that used to cost me 45 minutes of rereading produced a site-ready FAQ in about 70 seconds of total work.
A content brief commissions the writing; the FAQ is often the part visitors actually read, and it runs on the same bracketed-slot pattern. I pasted the full service description into Claude with every personal detail swapped for a placeholder first, and hit enter.
Prompt to copy
You are a technical writer. I will paste a document below. Read it and generate a FAQ section for a website. Rules: (1) write eight to twelve questions a real first-time reader would ask, in plain English; (2) keep each answer to two to four sentences; (3) lead with the most important question; (4) flag any question where the document gives an unclear or incomplete answer – mark it [GAP: source needed]; (5) do not invent answers – if the document does not cover something, use the gap flag, never guess. Here is the document: [paste it].
Claude returned the FAQ in 60 seconds.
Two answers carried a [GAP: source needed] flag where the policy genuinely never said what happens after cancellation, so I filled those from the real process in 10 minutes of review, and the whole task landed under 20 minutes against 45 by hand.
14. PDF or Long Report Summary Prompt
A chatbot can invent a number that looks right. That is a hallucination: a confident, well-formatted figure with nothing behind it. The longer the document, the more chance it misreads a figure, so every number must be checked against the original page.
The old way was 90 minutes of reading and note-taking, plus a second pass to confirm every figure.
Now I paste the report into Claude, run the prompt below, and get a three-sentence executive summary, five findings under 25 words each, two or three recommended actions and a numbers table in 90 seconds.
Prompt to copy
You are a senior analyst. I will paste a report or document below. Summarize it for a busy [executive / manager / client] who has five minutes. Format: (1) three-sentence executive summary; (2) five key findings as bullet points, each under 25 words; (3) two to three implications or recommended actions; (4) one table of the most important numbers, if the document contains them. Flag any finding that seems surprising or that I should verify against the original before sharing. Here is the document: [paste it].
Then I spend 15 minutes cross-checking each figure in that table against the source and reading the flagged findings. Shared with a manager, the summary is the AI's wording and my numbers, and the check is what makes that split safe.
15. Competitor Research Digest and Gap Table
Prompt to copy
Act as a strategic analyst. I will paste notes, screenshots or excerpts from [number] competitor websites or materials below. Summarize them into a competitive digest. Format: (1) a comparison table with one row per competitor and columns for: positioning, target customer, top three claims, pricing model (if visible in my notes), and one notable weakness; (2) two to three strategic observations about gaps or opportunities; (3) one recommendation for how I should position differently. Do not invent facts not in my notes. Flag any column where my notes are thin and I need more research. Here are my notes: [paste them].
That prompt turns scattered notes into one table plus a positioning call.
Summarizing a single report pulls out its points; this asks Gemini to hold four competitors side by side, and thin notes produce a thin table, which the gap flags make visible before anyone sees it.
Four competitors took 45 seconds to produce and 20 minutes to review, against 120 minutes of synthesis by hand. The flagged columns told me exactly which competitor pages to revisit, so the 20 minutes went to real research, not formatting.
16. Meeting Notes Converted to Action Items
After a project meeting, the worst job is the write-up. You remember four decisions, you know who agreed to what, and then you sit down to type it. Mine took 40 minutes each time, with a different format almost every week.
One detail caused most of the mess. My notes used three versions of one colleague's name. The AI assigned a task to whichever version appeared most often, and the owner column came out wrong. I now standardize names before I paste anything.
Prompt to copy
You are a project coordinator. I will paste raw meeting notes below. Extract and organize: (1) a decision log – each decision made, stated in one sentence; (2) an action item table with columns: task, owner (use the name from the notes), due date (use dates mentioned or write [DATE TBD]), and priority (High / Medium / Low); (3) a two-sentence summary of the meeting I can paste into the team channel; (4) any open questions that were raised but not resolved. Do not add items not mentioned in the notes. Here are the notes: [paste them].
The output is a decision log, a prioritized task table, a two-sentence channel summary and an open-questions list. The AI answers in about 20 seconds and I spend 8 minutes checking owners. The 40 minutes become under 30, and the format never drifts.
📋 How to Build Your Prompt Library in One Week
Day 1 — Audit your recurring tasks
List every task you repeat weekly or monthly and note how many minutes each one actually costs you.
Day 2 — Pick the three highest-cost tasks
Find those three rows in the table above and run each prompt once with real material from a recent project.
Day 3 — Review and save the outputs
Compare each AI draft to what you would have written, note what needed fixing, and save the refined prompt to your notes app.
Day 4–5 — Add five more prompts
Work through the next five tasks on your list, one at a time, saving each prompt as a named template.
Day 7 — Run the weekly planning prompt
Use item 20 to plan next week using the library you just built, so the habit starts on day one of week two.
17. Sanitize-Before-You-Paste Privacy Prompt
The moment an identifiable name, account number or confidential label goes in with them.
On free and personal plans, chats may train the company's models unless you switch that off in settings. So no passwords, client files, medical or financial records go in first.
The catch is the manual scrub most people skip. Finding every name and figure in a five-page thread was a 30-minute find-and-replace I kept postponing, which meant it never happened. Run this Claude prompt first, then paste the redacted result into any other prompt here.
Prompt to copy
Before I paste anything into this chat, help me strip it of sensitive information. I will paste a document, email thread or set of notes. Your job: (1) identify every piece of personally identifiable information – names, email addresses, phone numbers, account numbers, dates of birth, physical addresses; (2) identify any text marked confidential, proprietary or subject to an NDA; (3) produce a redacted version where each removed item is replaced with a bracketed label such as [CLIENT NAME], [ACCOUNT NUMBER] or [DATE]. Do not summarize or change anything else – only redact. I will use the redacted version in my next prompt. Here is the document: [paste it].
The ledger: By hand 30 minutes; AI output 30 seconds; my review 10 minutes. It turns a 30-minute scrub people skip into a 30-second one they do, and the redacted file still works in every prompt after it.
18. One-Page Project Brief from Rough Inputs
Do not hand a project brief to AI when the scope and timeline do not exist yet.
ChatGPT will flag a vague [timeline] or [budget] slot rather than paper over it, which is deliberate. The flags expose the thinking you have not done.
Get those constraints real and specific before you run anything.
Once the inputs are solid, the same prompt turns goal, scope, stakeholder roles, timeline and budget into a one-page brief in seconds, with a short risks section at the end.
By hand that organizing took me 60 minutes. A redacted document is safe to feed in as background.
Prompt to copy
Act as a project manager. Write a one-page project brief from the information I give you. Project name: [name]. Goal: [one sentence – what success looks like]. Background: [two to three sentences of context]. Scope – included: [your list]. Scope – excluded: [your list]. Key stakeholders: [roles, not names]. Timeline: [start date] to [end date]. Budget: [amount or TBD]. Format: use the above as section headers; keep each to three to five bullet points or sentences; end with a risks section listing two to three things that could derail the project. Flag any section where my input is too vague to write a useful brief.
The ledger: By hand 60 minutes; AI output 30 seconds; review and edits 12 minutes. About 45 minutes back, plus a list of the decisions you still need to make.
19. Case Study Draft from Client Project Notes
I timed one case study draft at 120 minutes. Three headlines, a 100-word challenge, a 150-word solution, a 100-word results section and a pull quote still needed a second full draft before the structure held.
Prompt to copy
You are a B2B case study writer. I will give you notes from a completed client project. The client details I provide are already anonymized. My notes: [paste them]. Target reader: [describe the prospect type this case study is meant to persuade]. Format: (1) three headline options in the shape '[Client type] achieved [result] in [timeframe]'; (2) challenge section, 100 words; (3) solution section, 150 words describing what we did step by step; (4) results section, 100 words, leading with the strongest metric; (5) a pull quote in the client voice – mark it [DRAFT: needs client approval before use]. Flag any result claim where I have not supplied a specific number.
Claude returned it in 60 seconds. The pull quote carries a [DRAFT: needs client approval before use] tag, and publishing it unsigned is my error, not the AI's.
Every result claim without a real number gets flagged; a vague result weakens the page. Review and sign-off: 25 minutes.
💡 Tip: Supply at least one hard metric — a percentage, a dollar amount, a number of days — in the results slot. Without it, the output reads like every other case study.
20. Weekly Planning and Daily Priority Prompt
Sunday planning now takes under 15 minutes and produces a task list sorted High, Medium and Low by deadline and impact.
It adds a morning-and-afternoon plan with deep work protected in the mornings, a short list of tasks to delegate or drop, and a three-sentence weekly intention I read each morning.
I paste my task list and calendar into ChatGPT using a prompt that opens "Act as an executive coach and productivity strategist."
Prompt to copy
Act as an executive coach and productivity strategist. I will give you my task list, calendar commitments and any notes about my energy or constraints for this week. Your job: (1) categorize every task as High / Medium / Low priority based on deadline and impact; (2) suggest a daily plan assigning tasks to mornings or afternoons – put deep work in the morning; (3) flag tasks I should delegate, delete or defer, with a one-line reason for each; (4) identify any scheduling conflict or overloaded day; (5) write a three-sentence weekly intention I can read each morning. Here is my week: [week]. Use honest time estimates for every task and flag any day where the total hours exceed my capacity.
It asks for all five parts, including any scheduling conflict or overloaded day. The AI answers in 30 seconds and my review takes 10 minutes.
That flag is the part that keeps the plan alive. My old Sunday version fell apart by Tuesday because I estimated everything too short. The AI inherits whatever times you feed it, so optimistic estimates in the [week] slot produce an equally unrealistic week.
Read the capacity warning first and cut a task before you accept the plan. Then paste the calendar back in with honest times. Here the plan lasts past Tuesday because the overloaded days are spotted before Monday, not on Wednesday afternoon.
Frequently Asked Questions
Do I need a paid plan to use these prompts?
Not to start. All 20 prompts work on the free tiers of ChatGPT, Claude and Gemini. Free tiers have usage limits that change, so check each product's pricing page before assuming you need a paid plan. Claude Pro and ChatGPT Plus both start at $20/month in the US as of October 2026 if you hit those limits quickly.
How do I know if the AI output is accurate enough to use?
Every prompt in this library includes a flag instruction — [CHECK], [WRITER TO SOURCE] or [GAP: source needed] — that marks claims the AI cannot verify. Read those flags first. Every number must be checked against the original source before anything goes live or is shared with a manager or client.
Can I use these prompts for any industry?
Yes. The bracketed slots are the key — replace [industry or niche], [type of company] and similar slots with your specific context and the prompt recalibrates its output to your situation. The newsletter prompt works for a law firm's client update as well as a product company's weekly email.
What happens if the AI output sounds nothing like my voice?
Paste a paragraph from a past piece of your own writing into the [brand voice] or [voice notes] slot of the relevant prompt. One concrete example of your voice outperforms any adjective like 'conversational' or 'professional.' Read the output against that same past piece before using it.
Is it safe to upload documents to Claude or ChatGPT?
Run the sanitize prompt in item 17 first. It strips names, account numbers and confidential labels and replaces them with bracketed placeholders. Use the redacted version in every other prompt. On free and personal plans, check each tool's privacy settings and switch off model training on your chats if the option is available.
How long does it actually take to get through the AI's output and review it?
The review time for each prompt is in the table at the top of this article — it ranges from five minutes for a short rejection email to 25 minutes for a case study. The AI's output time is seconds to about 90 seconds. The slow part is always the person reviewing, not the AI generating.
Find the Row That Hurts Most and Run It Once Today
Scroll back to the table at the top. Find the task where the 'By Hand' number hurts most — the 120-minute newsletter, the 90-minute social calendar, the 120-minute case study. Run that one prompt today with real material.
Compare the AI draft to what you would have written. That first experiment, not reading about it, is how the habit forms. Save the refined prompt to your notes app by name and it compounds from there.
Before you start: check the free-tier limits on ChatGPT and Claude at each product's pricing page. You may not need a paid plan for months.