Most “best AI prompts” lists are untested one-liners like “Act as a marketing expert.” This one is different: I ran all 12 prompts below in a fresh Claude chat with no memory and no setup, and I’m showing you what actually came back.
Every entry has four parts:
1. The prompt, ready to copy and paste.
2. Why it works: the technique you can steal for your own prompts.
3. What Claude gave back: a trimmed excerpt of the real output.
4. Remix it: how to adapt it to your own life or business.
Swap the details in the prompts (numbers, cities, your niche) for your own. The more specific you are, the better Claude gets. That is the whole secret, and you’ll see it in every example.
Part 1: Think better and decide faster
Prompt 1: The Pre-Mortem
I’m deciding whether to quit my stable corporate job (₹18 LPA, product manager) to go full-time on my side business — a D2C cold-brew coffee brand doing ₹2.5 lakh/month revenue with ~30% margins after 8 months.
Run a pre-mortem. Imagine it’s 18 months from now and this decision failed badly. Write the 5 most likely reasons it failed, ranked by probability. For each: the early warning sign I’d see in month 2-3, and one cheap action I can take NOW to prevent it. Then tell me the single number I should hit before quitting. Be blunt, not motivational.
Why it works: Asking “should I do it?” gets you cheerleading. Asking Claude to assume failure and work backwards gets you risks. The real numbers and the forced structure (warning sign, cheap fix, one number) make the answer usable.
What Claude gave back: It opened with maths I hadn’t asked for: ₹2.5L at 30% margin is about ₹75k a month, against roughly ₹1.15L take-home salary, so quitting now is a 35–40% pay cut.
• #1 failure: the margin wasn’t real once ad spend, returns and cold-chain losses were counted. Fix now: build a per-order P&L that includes a ₹50k salary for someone to replace you on ops.
• #2: growth stalled at the edge of your network. Warning sign: 60-day repeat rate under 25%.
• #4: seasonality. Cold brew dips in monsoon and winter, so quit in April and month 3 is July.
• The number: ₹1.2L a month in owner-takeable profit for 3 straight months, with at least one in the off-season.
Remix it: Use it for any big call: a hire, a move, a product launch, a new city. Always include your real numbers.
Prompt 2: Steelman the Other Side
I believe remote work is better than office work for most knowledge workers. Steelman the opposite view — make the strongest possible case against my position, the way its smartest defender would. Then tell me which of your arguments I’d struggle most to answer, and what evidence would actually change my mind.
Why it works: AI tends to agree with you. This prompt makes disagreement the task, then asks Claude to point at your weakest spot and name what evidence would settle it.
What Claude gave back: Five arguments, each tied to real research. The strongest: juniors pay the hidden cost of remote work, because tacit learning (overhearing a senior handle a client) disappears, and they can’t measure what they never learned. It cited the Microsoft study of 61,000 employees showing remote networks became more siloed. Its verdict: the honest version of my view is “remote is better for experienced workers,” which is narrower than “most.”
Remix it: Paste your business plan, pricing decision or a strong opinion you’re about to post publicly. Find the hole before your audience does.
Prompt 3: The Explanation Ladder
Explain how UPI payments work in India at three levels: (1) to a 10-year-old in 3 sentences, (2) to a college student in one paragraph, (3) to a fintech engineer with the actual flow — who talks to whom, in what order. End with one thing most people get wrong about it.
Why it works: Three levels in one answer lets you climb from the intuition to the real mechanics. The “what people get wrong” ending pulls out the insight most explainers skip.
What Claude gave back: A 10-year-old version with banks talking through a central “post office” (NPCI), then a six-step engineer flow from the payer app to NPCI to both banks to settlement. The ending was the best part: money doesn’t actually move between banks instantly. You’re credited instantly, but banks settle with each other later, in net batches through their RBI accounts.
Remix it: Use it on any topic you need to understand or teach: GST, the stock market, how Instagram’s algorithm works. Level 1 often makes a great reel script.
Part 2: Write content people stop for
Prompt 4: The Psychology Hook Generator
I make Instagram reels about personal finance for Indian millennials. Write 10 hooks (first 3 seconds, spoken) for a reel on “why your SIP isn’t making you rich.” Each hook must use a different psychological trigger — label the trigger. No clickbait lies, no “Did you know”. Then pick the top 2 and explain why they’d stop a scroll.
Why it works: “Different trigger for each” stops Claude from writing ten versions of the same line. Banning “Did you know” and clickbait removes the two most overused openers. Asking it to pick a winner gives you a decision, not just a list.
What Claude gave back: Ten labelled hooks covering loss aversion, curiosity gap, contrast, future regret and more. Its top pick, a contrast hook: “Two people, same ₹10,000 SIP. One ends up with more than double the corpus. The only difference is a 10% yearly top-up.” Without being asked, it checked the maths (assuming 12% returns) and told me to put “illustrative, assumes 12% p.a.” on screen.
Remix it: Change the niche and topic. Then ask for 3 more hooks in the style of the winner.
Prompt 5: The Ruthless Editor
Act as a ruthless editor at a top magazine. Here’s my LinkedIn post:
“I am super excited to announce that after a lot of hard work and dedication, our team has successfully launched our new product. It has been an incredible journey filled with ups and downs. I want to thank everyone who supported us. We believe this product will add a lot of value to our customers. Stay tuned for more updates!”
First, diagnose exactly what’s wrong in a bulleted list (be specific, not polite). Then rewrite it in under 80 words so a stranger would actually stop and read it. Use [brackets] for details I need to fill in.
Why it works: Diagnosis before rewrite teaches you what to fix next time. The brackets stop Claude from inventing fake facts about your product.
What Claude gave back: Nine problems, including “the product is never named,” “’add a lot of value’ is a non-claim,” and “you hint at a story and then hide it.” The rewrite opened on a customer’s pain instead of the founder’s excitement:
[Number] months ago, a [customer type] told us: “[exact painful quote].” Today we’re launching [Product name] to fix that.
It then added one proof number, one honest lesson and a clear ask.
Remix it: Paste your bio, your last caption or your website’s homepage.
Prompt 6: One Idea to a Full Carousel
Turn this one idea into a 7-slide Instagram carousel: “Most people use AI like Google. The winners use it like an employee.”
For each slide give: the headline (max 8 words), the body text (max 25 words), and a one-line visual direction. Slide 1 must be a hook, slide 7 a CTA to save and follow. Make each slide end so the reader has to swipe.
Why it works: Word limits keep slides readable on a phone. “End so the reader has to swipe” builds a loop into every slide, which is what drives carousel completion.
What Claude gave back: A complete carousel ready for Canva. Slide 1: “You’re using AI wrong. Most people are.” Slide 4 introduced a 4-part brief (Role, Context, Task, Standard). Slide 5 showed a weak prompt next to a strong one. Every slide ended on a cliffhanger such as “Here’s the exact brief →” or “One last thing…”
Remix it: Feed it the best idea from any of your Substack posts, and you have this week’s carousel.
Part 3: Learn skills and level up your career
Prompt 7: The 80/20 Learning Sprint
I want to learn video editing in DaVinci Resolve well enough to edit my own YouTube videos. I have 30 minutes a day for 3 weeks. Build me a day-by-day plan using the 80/20 rule — only the skills that give the most results. Each day must end with a tiny project I can finish. Tell me what to deliberately ignore.
Why it works: The time limit (30 minutes, 3 weeks) forces Claude to prioritise. The “tiny project” rule turns watching into doing. “What to ignore” is often the most valuable part of the answer.
What Claude gave back: A 21-day plan in three phases: cutting, then sound and look, then your real workflow. Day 3: remove every “um” from a 2-minute clip. Day 21: a published video and a saved project template. The ignore list was gold: the Fusion page, node-based grading, LUTs and fancy transitions, because “your viewers notice bad audio and slow pacing, not color science.”
Remix it: Swap in any skill (Excel, Canva, public speaking, German) and your real time budget.
Prompt 8: The Strict Professor
Here’s how I’d explain inflation to a friend: “Inflation is when prices go up because the government prints too much money.” Find every gap, oversimplification, or error in my explanation, like a strict economics professor would. Then give me a better explanation that’s still under 60 words, and 2 questions I should be able to answer if I truly understand it.
Why it works: Explaining something in your own words, then having it torn apart, is the fastest way to find what you don’t actually understand. The test questions show whether the gap is fixed.
What Claude gave back: Six corrections. Inflation is a sustained rise in the general price level, not one product getting dearer. Most money is created by commercial banks lending, not by government printing. After 2008, huge money creation in the US and Japan produced little inflation. Its test question: “If petrol prices double while most other prices fall, is that inflation?”
Remix it: Use it on anything you’re studying: an exam topic, a concept in your industry, or a claim you’re about to make in a reel.
Prompt 9: The Negotiation Rehearsal
I have a salary negotiation call tomorrow. I got an offer of ₹14 LPA for a Senior Content Strategist role; I want ₹17 LPA. Play the HR manager, who is friendly but has been told the budget is tight. Write out a realistic 8-exchange dialogue where I negotiate well, then a short breakdown of the 3 moves that worked and the exact sentences I should memorise.
Why it works: A full script shows you how the conversation moves, not just tips. “Budget is tight” makes the practice realistic. “Exact sentences to memorise” gives you something to use under pressure.
What Claude gave back: A realistic dialogue that ended at ₹16 LPA plus a ₹60,000 joining bonus and a written six-month review. The three moves: anchor on market data and scope, widen the pie when base pay is stuck, and use a conditional close: “If we can get to [package], I’m ready to accept this week.” Bonus tip: after you say your number, stop talking.
Remix it: Use it for a rent negotiation, a client price increase or a tough conversation with your manager. Follow up with: “Now play a tougher HR manager.”
Part 4: Run your business like a pro
Prompt 10: The Customer Objection Simulator
I’m launching a ₹999/month online fitness coaching program for working women aged 28-40 in Indian metros. Simulate 3 very different potential customers who are on the fence. For each, give a name, one-line background, and their honest internal objection in their own voice (not polite). Then tell me which objection is most dangerous to my business and write the exact landing-page line that answers it.
Why it works: “In their own voice (not polite)” gets you what customers think, not what they say in surveys. Asking which objection is most dangerous forces a priority, and the landing-page line turns the insight into copy.
What Claude gave back: Three customers: Priya, a Bengaluru PM with a 4-year-old who had quit three fitness apps; Neha, who thinks YouTube is free; and Sunita, worried about PCOS and knee pain. It picked Priya’s as most dangerous because she already expects to fail, which hurts both conversion and retention. The line it wrote: “Built for the weeks you’ll want to quit.” Then it added a caution: only use that line if the WhatsApp check-ins it promises are real features.
Remix it: Run it before you set a price, write an ad or build a sales page.
Prompt 11: Messy Notes to SOP
Turn my messy notes into a clean SOP a new hire could follow on day one:
“ok so when client sends brief, check if budget mentioned, if not ask. then make a folder in drive w client name + month. copy template deck. sometimes we forget to add to tracker sheet, pls do that. send kickoff mail within 24 hrs, cc account manager. if its a rush job (<5 days) tell rahul first. after kickoff fix call slot.”
Format: numbered steps, owner for each step, a ‘common mistakes’ box, and a definition of ‘done’. Flag anything ambiguous in my notes as a question.
Why it works: You give it your real voice-note mess, and it gives back a process. “Flag anything ambiguous” is the key line: Claude asks instead of guessing, so your SOP doesn’t quietly invent rules.
What Claude gave back: A 7-step table with owners and deadlines. It moved the rush-job check to step 2, so Rahul hears about it before any work starts. It added a “definition of done” checklist and 8 sharp questions, such as “Is a rush job under 5 calendar days or working days?” and “Should we wait for the budget before sending kickoff?”
Remix it: Record a 2-minute voice note explaining any task you repeat, transcribe it and paste it in.
Prompt 12: The Cold Email That Gets Opened
I need to cold email the founder of a fast-growing Indian D2C skincare brand to pitch my video production agency. She gets 200+ emails a day. Write 3 versions under 90 words each, using three different strategies (label them). Subject lines must be under 6 words. Then tell me which you’d send and the one mistake that would make all three fail.
Why it works: Three different strategies let you test approaches, not just wording. The “one mistake that makes all three fail” question gets Claude to name the blind spot most people miss.
What Claude gave back: Three emails: a specific observation, a give-first offer and peer proof. It would send the give-first version, a 20-second reel made on spec for her hero product, because “a finished video about her product is hard to ignore.” The mistake that kills all three: generic personalisation. Every bracket needs a real, checkable detail, and you should never invent metrics.
Remix it: Change the industry and your offer. Then ask: “Now write the follow-up I send 4 days later if she doesn’t reply.”
The 5 rules behind every prompt here
Look back at the 12 prompts and you’ll see the same five moves. Use them and you won’t need anyone’s prompt list, including mine.
1. Give real numbers and context. “₹18 LPA, ₹2.5 lakh/month, 30% margins” got sharper advice than “should I quit my job?” ever would.
2. Shape the output. Word limits, slide counts and “numbered steps with owners” tell Claude exactly what done looks like.
3. Ban the boring default. “No ‘Did you know’”, “be blunt, not motivational” and “not polite” stop the generic answer before it starts.
4. Make it choose. “Pick the top 2,” “which is most dangerous” and “which would you send” turn a list into a decision.
5. Ask for the blind spot. “What do people get wrong,” “what would change my mind” and “what mistake makes all three fail” pull out the insight you didn’t know to ask for.
One more habit: treat the first answer as a draft. Reply with “too generic, make it punchier” or “now do it for a tougher audience.” The second or third round is usually where it gets great.
What’s next
This is part one of a series where I test AI tools so you don’t have to. Next up, I’ll run prompts like these through other popular AI tools and show you the real outputs side by side.
If one of these prompts saved you an hour, here’s how to get more:
• Subscribe so the next tested guide lands in your inbox.
• Follow me on Instagram @kevinsingh.me for one tested prompt every week, with the real output on screen.
• Reply or comment with the prompt that worked best for you, or a task you want me to test next.