The Cynical PM's Prompt Playbook
Mastering the Art of Robot Whispering
Welcome to 2025, where your professional value now depends on your ability to sweet-talk a text box. Remember when AI was supposed to make your job easier? Surprise: it's actually created an entirely new skillset you need to master just to keep pace with Chad, the Growth PM who won't shut up about "crushing KPIs with AI automation" while high-fiving himself in Slack threads.
As if your job description wasn't already a clusterfuck of impossible demands: understand the market, wrangle data, read customers' minds, lead without actual authority, communicate like a TED speaker crossed with Hemingway. Alas, the universe has conspired to add one more essential skill to your bloated professional repertoire: becoming an AI whisperer.
Google just dropped a 69-page white paper (nice) on prompting because apparently talking to computers requires more instruction than raising actual human children. I've distilled their academic masturbation session, plus my own insights from spending way too much time with AI, into this guide that will make you look 800% better than your colleagues still typing "make good deck" into ChatGPT. Let's dive into this fresh circle of professional hell, shall we?
How to Out-Prompt your Colleagues: A TL;DR
This unholy manifesto contains four essential techniques to manipulate language models into doing your bidding before they inevitably take your job:
Contextual Prompting: Giving AI the background information it needs. When you say "I need recommendations for a family vacation," you'll get Disney World. When you say "I need recommendations for a family vacation with two ADHD toddlers, a partner who hates crowds, during a government shutdown, with a budget of exactly $1,200," you'll get something actually fucking useful.
Persona Prompting: Want advice from a CMO without the seven-figure salary? A lawyer minus the billable hours? Steve Jobs sans the turtleneck and god complex? Just tell the AI to engage in some digital cosplay.
Step-by-Step Prompting: Breaking tasks into discrete stages forces models to engage with specific sub-problems rather than pattern-matching to the most statistically likely (and usually mediocre) output.
Meta-Prompting: The ultimate intellectual laziness: making AI write its own prompts. Why learn prompt engineering when you can just tell the AI to prompt itself?
Why Prompting Matters
Prompt design = UX for LLMs. Great prompts guide/focus the machine’s “attention.”
Prompts shape what LLMs say, how they say it, and how useful it is. Garbage prompts = garbage outputs.
You're basically programming with English. Consider yourself a digital Hogwarts student; prompts are your spells. It’s time to get good or risk detention with Professor Snape.
Technique 1: Contextual Prompting
What it is: The revolutionary concept of providing relevant background information, something human conversations have managed since the dawn of language, but apparently deserves a fancy technical term in AI. It means handing the model the same briefing points you’d give a human: purpose, audience, medium, length, and tone.
“Write me an email asking for funding approval.”
-OR-
“Write an email to persuade senior finance execs (new to tech) to approve our funding, use a formal and reassuring tone, keep it short, and use bullet points.”
Why it works: LLMs succeed by pattern-matching against mountains of text. Clear context narrows the haystack: it steers the model toward examples that match your goals, audience, and format, trimming fluff, lowering hallucination risk, and reducing back and forth.
Getting Contextual
Here are the types of context to consider providing in your prompts:
🎯 Purpose/Goal
Intention: Inform, persuade, educate, entertain, summarize.
Outcome desired: Actionable recommendations, conceptual overview, tactical details, inspirational tone, humor.
Example: This presentation is for the board of directors and is intended to persuade them that we had a successful Q2 while approving additional funding for Q3.
🎩 Audience
Role/Seniority: Executive, senior leadership, mid-level managers, entry-level, customers.
Audience Knowledge Level: Expert, intermediate, novice.
Interest/Concerns: Strategic priorities, ROI, risk, competitive positioning, brand impact, operational efficiency.
Outcome desired: Actionable recommendations, conceptual overview, tactical details, inspirational tone, humor.
Example: The audience is senior executives who are new to technical concepts and expect actionable recommendations in a concise format.
✉️ Medium
Medium: Slide deck, email, report, social media post, newsletter article, internal wiki, Slack/Teams message.
Length: Concise bullet points, detailed paragraphs, brief email, extensive analysis.
Example: The output should be in the form of a section header that is also the key takeaway point, and then should present supporting context in bullet points.
💬 Style/Tone
Style: Formal, informal/conversational, technical, educational, promotional, humorous.
Voice: Analytical, provocative, reassuring, irreverent, neutral.
Example: Please use a formal tone that sounds reassuring and inspires confidence.
Contextual Prompting works dramatically better than letting models operate in a vacuum of assumptions, almost as if understanding context matters. Revolutionary concept, I know. While seemingly obvious, the majority of humans are too lazy to do this, and then bitch about the hot garbage that AI spits back.
Technique 2: Persona Prompting
What it is: You assign the model a specific persona or role using your prompt.
Why it works: This helps the model filter relevant details aligned with a role’s priorities, concerns, and expertise. There are many ways to approach persona prompting:
🧑💻 Role
Useful for: Framing context and receiving more relevant answers.
Examples:
Software Product Manager: For when you need to sound like you know what the fuck you're doing in the meeting you're unprepared for in 15 minutes.
Product Designer: When you to create a UI that makes your IRL design team question their career choice.
User Researcher: When you need to develop interviews, survey guides, or research approaches but are too poor to afford the flesh and blood version.
Copywriter: For creative language work and making your copy suck less.
CMO: All the strategic marketing puffery without the seven-figure compensation package and ego management.
Your Customer: Because who has time to talk to actual customers?
Prompts:
Embody a user researcher who specializes in identifying unspoken pain points and analyze our current workflow to identify the three biggest frustrations users won't explicitly tell us about but are definitely experiencing.
Adopt the perspective of a Head of Product who's shipped features used by millions and critique our roadmap with a specific focus on which items should be killed immediately because they're vanity projects rather than focused on customer needs.
Channel the viewpoint of a customer who's evaluated all our competitors and is struggling to justify our premium pricing tier. Articulate the specific objections they'd have so we can preemptively counter them in sales materials.
Act as a senior UX copywriter with 15+ years of experience in conversion optimization who has A/B tested over 500 email capture forms. Generate 5 ways to ask users for their email addresses that make them feel like they are receiving reciprocal value when they're actually being funneled into our marketing ecosystem.
Embody a veteran CMO who's successfully repositioned three legacy brands during market downturns. Review our current positioning and provide three strategic shifts that would make Gen Z hate us less without alienating our boomer revenue base.
🧐 Expert Advisor
Useful for: Getting an expert opinion from a specialist minus the hassle of asking a coworker or shelling out for a consultant.
Examples:
Startup Advisor: When you need someone to critique your wildly optimistic pitch deck.
Corporate Counsel: To save you from tanking the company via your unlawful decisions. See also: creating or reviewing contracts and agreements.
Career Coach: To translate "you're a self-absorbed asshole" feedback into actionable steps that don't require fundamental personality changes you're unwilling to make.
Jungian Dream Analyst: To decode the high strangeness of your dreams.
CFO: When you need financial advice from someone who won't laugh directly in your face when you show them your revenue projections.
Prompts:
As a brutally honest startup advisor who's seen over 1,000 failed startups, review our pitch deck and explain why we're most likely to end up as a cautionary tale on Hacker News rather than a TechCrunch success story. Focus specifically on the delusional market size calculations and the competitive advantage we think is unique, but actually isn't.
Embody our corporate counsel with 15+ years of tech industry experience and explain the specific ways our user data collection practices are putting us at catastrophic legal risk across different jurisdictions, ranking them from 'aggressive but defensible' to 'executives might actually see jail time.
Act as a brutally honest career coach who specializes in translating feedback into actionable steps. Please analyze my performance review and create an actionable development plan that addresses major feedback areas while capitalizing on my strengths.
Pretending you are a Jungian Dream Analyst, please help me analyze the following dream, asking any follow-up questions you need to create a more insightful and complete analysis.
🎭 Creative Role
Useful for: Creative ideas, storytelling, fresh perspectives that break conventional product approaches, humor.
Examples:
Ad Agency Creative Director: For that cocaine-fulled creative energy without the rehab bills.
Game Designer: Because who understands
manipulationgamification better than people who design dopamine loops?Disney Imagineer: For that magical ability to make people forget they're being monetized.
Screenwriter: Masters of emotional tension and narrative structures.
Stand-up Comedian: When you need uncomfortable truths wrapped in plausible deniability.
Pixar Story Architect: When you need to add a dash of panache or storytelling magic to your boring pitch.
Prompts:
As a Game Designer who's shipped multiple successful titles, please identify three engagement mechanisms to make our onboarding experience as engaging as Candy Crush.
Please channel a Disney Imagineer who's worked on immersive park experiences and redesign our onboarding to create the same emotional resonance that makes parents spend $15 on Mickey-shaped ice cream.
Adopt the perspective of a Stand-up Comedian who's survived ten years of brutal club circuits. Please draft error messages that make users laugh instead of switching to our competitors when our product inevitably fails.
Embody an Emmy-winning HBO Writer and create a tutorial narrative with enough tension, character development, and payoff that users will actually complete it instead of immediately Googling 'how to skip tutorial in [product name].’
As a Pixar Story Architect, please craft a customer journey that emotionally resonates with users through vulnerability and triumph so they form an irrational attachment to our product.
📚 Book, Framework, or Blog
Useful for: Applying a particular approach or framework to a problem. We’ve all been there, you read books to apply their knowledge in real life, only to have the following play out: Spend your rare down team reading a book → Encounter real-life scenario in which the book could help you → Fucking forget whatever practical knowledge you theoretically acquired.
Examples:
48 Laws of Power: When you need to manipulate colleagues like a Game of Thrones court schemer but lack the knowledge or tactical patience.
Laws of Human Nature: When you need to understand why your teammates are sabotaging your project, but lack the emotional intelligence to figure it out yourself.
Elephant in the Brain: For explaining why your customers don't give a fuck about your "mission" and just want status signals they can post on Instagram.
Influence: For getting your way with colleagues or customers.
Multipliers: For leadership advice that actually doesn’t suck.
The Hard Thing About Hard Things: For non-bullshit leadership advice and how to recover when you’ve fucked something up.
Jobs-to-be-done: When you need to explain that people aren't buying your product for the reasons marketing thinks they are, but for entirely self-interested purposes you've been ignoring.
Behavioral Science: When you need to weaponize Human OS against the zombies that run it.
Stratechery by Ben Thompson: When you need to sound like you understand complex platform economics beyond just saying "capitalism gonna capitalize."
Prompts:
Through the lens of Jobs-to-be-done, analyze why our customers are actually buying our productivity software despite claiming it's for 'team collaboration'.
Leverage the behavioral science concept of scarcity bias to redesign our pricing page to create artificial urgency.
Using Cialdini's six principles of influence, craft a strategy to convince our engineering team to prioritize my feature requests when they're already drowning in tech debt, haven't slept more than four consecutive hours in a month, and the CTO just announced another "visionary pivot."
What does Liz Wiseman’s book Multipliers say about leading via establishing high standards?
🎓 Specific Person
Useful for: Applying a specific expert’s framework or secret sauce to a problem. Think: intellectual fan fiction meets corporate utility.
Examples:
Rory Sutherland: For a persona focused on creativity and behavioral science, who leans into unconventional insights and counterintuitive ideas.
Erika Hall: For a master researcher with brutal honesty and a no-bullshit approach to data collection.
Marie Kondo: For when you need someone to tell you to kill 90% of your feature bloat but lack the courage to say it yourself without invoking a Japanese woman as your spiritual authority.
Harry Dry: For when you need a copy that actually sells shit instead of the masturbatory "brand voice" your marketing team created after a three-day offsite at a vineyard.
Steve Jobs: The PM equivalent of a basic bitch Instagram caption. For when you need to embrace your inner masochist and hear that your ideas are ‘complete shit’.
David Lynch: For when conventional product thinking is failing and you need to get weird.
Hunter S. Thompson: For the authentic experience of making product decisions while on a cocktail of hallucinogens and stimulants without the medical risks or legal consequences.
Prompts:
As Steve Jobs reviewing our product with the same brutal honesty he used in Apple product reviews, which aspects would you call 'complete shit' and demand we fix before launch?
Channel Erika Hall at her most brutally honest. As the author of 'Just Enough Research' who's built a career calling out bullshit research practices, review our latest customer survey questions and tell us exactly why they're leading, useless, and designed to confirm our existing biases rather than reveal uncomfortable truths. Then rewrite them so they'll actually give us data worth a damn instead of the self-congratulatory nonsense we're currently collecting.
Embody Harry Dry of Marketing Examples fame, the king of high-conversion direct response copywriting who's built his entire reputation on stripping away marketing bullshit and focusing on concrete, specific value. Review our current product landing page and tell us why our vague feature descriptions and meaningless corporate buzzwords are precisely why our conversion rate is in the toilet. Then rewrite our hero section copy with the same surgical precision you'd use in your newsletter, making it so specific and value-focused that even the most skeptical visitor would be compelled to sign up. And for fuck's sake, make it something a normal human being would actually say out loud rather than the corporate word salad we currently have plastered everywhere.
Channel Rory Sutherland at his most contrarian. What psychological hack could make our basic subscription tier feel premium without changing the actual features?
Taking Hunter S. Thompson's gonzo approach to product research, write a stream-of-consciousness narrative of our user's experience that captures the existential dread and confusion they're not telling us about in surveys.
🧙 Historic & Fictional Figures
Useful for: Getting sage advice from great minds of yore or your favorite fictional characters. See also: Indulging your embarrassing fandoms under the guise of professional development.
Examples of Entities You Can Digitally Resurrect/Summon:
Plato: When you need to dress up basic workplace politics as profound philosophical discourse instead of the petty bullshit it actually is.
Marcus Aurelius: For Stoic justifications of why your team should accept terrible working conditions without complaint, because "suffering builds character."
Albus Dumbledore: When you need manipulative leadership advice that sounds wise but actually involves withholding critical information from your team "for their own good.”
Leslie Knope: For channeling pathological enthusiasm.
Sherlock Holmes: When you need to cosplay as an intellectual genius who sees patterns others miss, rather than admitting you're just guessing like everyone else.
Chrisjen Avasarala: For when you need to combine ruthless strategic thinking with creative profanity that would get you fired if it came directly from your mouth.
Prompts:
Please pretend you are Plato, what guidance would you offer a product manager feeling overwhelmed by impostor syndrome?
Embody Marcus Aurelius writing a private Meditations entry after witnessing a catastrophic product launch. Please provide stoic counsel on coping with the failure and moving forward.
As Sherlock Holmes investigating our product's declining retention metrics, what seemingly insignificant details would you observe that everyone else has missed? Which 'elementary' insights would reveal the true cause that we've all been too stupid to notice?
Channel Chrisjen Avasarala at her most strategically profane. How would you advise me to navigate corporate politics where multiple VPs are trying to sabotage my product while maintaining plausible deniability?
Bonus Points: Use “descriptive adjectives” to make your role-based prompts even better
For bonus points, use adjectives and descriptions of the role. You could say:
Pretend you’re a product manager
-OR-
Pretend you’re a seasoned software product manager who has built multiple B2C products used by millions
This nudges the model to adopt a more experienced or nuanced tone, leading to more detailed, insightful responses. You’re giving a hint about the level of expertise you expect. Think of this like selecting attributes for a character in a video game or putting out a vibe that the AI tries to match.
Technique 3: Step-by-Step Prompting
What it is: Step-by-step prompting is essentially project management for AI—forcing these over-confident language models to think step by step show their godsdamn work instead of jumping straight to conclusions that make you question whether the singularity is just a Silicon Valley fever dream.
When it’s useful:
Complex tasks with multiple steps: “Generate 10 ideas → rank by opportunity → draft slide copy for top three.”
Tasks that depend on referencing research or data: “Review and summarize the attached customer research, then ask me which persona to focus on.”
Exploratory or Branching Paths: You’re not sure which direction you’ll take until you see the first output: “Idea generation → pick your favorite → create prototype → write PRD”
Example:
Step 1: Role → Please pretend you’re a seasoned B2C product manager with a track record of applying psychological frameworks to product development.
Step 2: Research → Please review the attached customer persona research.
Step 3: Generate Ideas → Generate 20 specific product improvement ideas that address conversion challenges.
Step 4: Evaluate → Develop a scoring system with the following components, using a 1-10 scale: Potential conversion impact, Implementation effort (10=lowest effort), Percentage of users impacted, Long-term strategic value.
Step 5: Prioritize → Score all 20 ideas using the framework.
Practical Tips for Step-by-Step Prompting
Name your steps: “Step 1: Outline → Step 2: Research → Step 3: Draft,” so you and the model can refer to them unambiguously.
Echo critical constraints in every step: “Maintain brand voice guidelines linked above.”
Save yer prompts: Store templated multi-step prompts in your notes app of choice; tweak only the specifics each time you prompt.
LLMs perform better with step-by-step prompting because it simplifies complex tasks, reducing cognitive load and the risk of errors. This approach mimics how actual humans solve problems—iteratively, with checkpoints, and the humility to revise when things go sideways—rather than pretending AI can magically understand your vague, contradictory requirements on the first try.
Technique 4: Have AI Write the Prompt
The ultimate prompt engineering lifehack isn't learning prompting at all—it's making the AI write its own fucking prompts. Why waste precious brain cells crafting the perfect prompt when you can just tell the AI to do it for you?
🪄 Make AI Fix Your Dogshit Prompts
Here's a meta-technique for when you're too intellectually bankrupt to craft a decent prompt but still want AI to save your ass:
Step 1) Write yer prompt: Vomit your half-assed prompt onto the page like the digital equivalent of drunk texting your ex at 2 AM
Step 2) Copy/Paste: Copy/paste this magic incantation ABOVE your garbage-tier first attempt:
Can you please help me create a refined prompt by analyzing the original prompt (see below) for core intent, subject domain, and implicit objectives? Evaluate the prompt's structure for clarity, context, and organization. Then enhance it by: adding necessary context and clarifications, implementing clear structural elements like bullet points or numbered steps; defining constraints, style preferences, and tone requirements; specifying output format and reasoning requirements; and including verification mechanisms. Finally, reconstruct an optimized version that maintains the original request while incorporating all improvements, and briefly explain the key enhancements you've made and their purpose.
Step 3) Use the new prompt: Use the improved prompt and take full credit for the results in your next meeting.
Shout out to Alena Panshina for sharing this one.
🔬 Deep Research AI Agents
Several AI tools, including ChatGPT and Gemini, have ‘Deep Research’ agents that use reasoning to synthesize large amounts of online information and complete multi-step research tasks for you. Consider this your personal research department that never sleeps, complains, or asks for credit.
Here’s how to get the most out of these tools:
Step 1: Write down what you want to know
Step 2: Ask the model you are using to write a prompt for that model’s Deep Research function
Step 3: Turn on Deep Research mode and copy/paste your prompt
Example: Can you please help me write a prompt for ChatGPT Deep Research? Here's what I want to know:
I work for PODS moving and storage, our email marketing wants to understand what strategies work best for open rates and click-through rates
We would like to know what works best for each phase of the customer journey - e.g., before quote, re-engagement after a quote, promotional sell
What emails are generally most effective - highly styled/designed or simple text?
What content and components should we prioritize for A/B testing with our emails?
Are there any learnings specific to the moving and storage industry?
⚡ No Code Builders like Lovable & Bolt
Step 1: Describe the product/website you want to build
Step 2: Ask AI to create a prompt
Step 3: Copy/paste into Lovable/Bolt/Replit/etc.
Example: Based on the product we've discussed creating in this chat, can you please help me create a prompt for Loveable, an AI that generates websites, to create a landing page for this product?
🖼️ Image Generation
For when you need AI to design the perfect custom doormat based on your maniac Chihuahuas.
Example: Hi, can you help me write a prompt for an AI image generator to take a picture of my Chihuahuas and turn it into a doormat? The doormat should say, "No need to knock, Sammich & Mochi already know you are here." I want it to be in the feel-good style of a 90s cartoon with bright, bold colors.
🎵 Audio Generation
For when you want to use products like Suno to create ridiculous songs for your offspring:
Example: Hi, can you please help me write a prompt for an audio generation AI tool so it will create a playful, imaginative pirate-themed song for my kids, Gibby (a 3-year-old boy) and Graham (a 2-month-old girl), in under 200 characters?
The Pirate Pancake Adventure gets a lot of play at home.
What's happening beneath the surface:
You're
exploitingleveraging the AI's training on prompt engineering literatureThe AI is ghostwriting instructions to itself
You get to feel clever while doing even less work
Rapid Fire Tips & Tricks
Say "Please," You Animal. A 2024 study found polite prompts improved AI accuracy by 9%. These digital entities respond better to basic courtesy than most of your coworkers.
Create Projects for Context. Stop one-off copy-pasting your company strategy docs. Tools like Claude let you create persistent projects where your AI can reference all your corporate bullshit without you repeating it every time.
Shop Around. Different models, different strengths. Claude might nail your creative brief while Gemini solves your coding problem. Dating multiple AIs isn't cheating; it's strategic promiscuity.
wrote a great article on this: An Opinionated Guide on Which AI Model to Use in 2025
Document Your Prompts. Keep a swipe file of what works. Your future self will thank you when you're not recreating the same prompt engineering masterpiece at 11 PM before a deadline.
Examples > Explanations. Show, don't tell. One good example is worth a thousand words of explanation.
Be Obnoxiously Specific. "Give me headlines" gets you garbage. "Give me 5 landing page headlines in sentence case that use action verbs and address the underlying pain points of project management overwhelm" gets you results.
Instructions > Constraints.. "Write an email whose audience is senior leaders at company xyz" works better than "Don't write an email that sounds like a Nigerian prince scam." Positive guidance beats negative constraints.
Make It Ask Questions. End your prompts with "Please ask any clarifying questions if needed." The AI will help you help it, which is meta-laziness at its finest.
Congrats, You're Now Fluent in Machine Manipulation
We're all frantically learning to communicate with machines because the alternative is watching Chad from marketing get promoted for the deck that Claude actually made. The game is rigged, but at least now you know how to play it.
The good news? This skillset isn't permanent. In five years, we'll either be communicating with AI through direct neural interfaces, or we'll be hiding in underground bunkers using these same techniques to convince the machines we're not worth harvesting for nutrients.
Until then, may your prompts be precise, your outputs be usable, and your colleagues remain blissfully unaware that your "creative breakthrough" came from typing "channel Hunter S. Thompson experiencing an existential crisis while reviewing our Q3 marketing strategy" into a text box at 2 AM. 🖤
BONUS CONTENT
🫨 The weirdest role prompts that might still be useful
Cult Leader: For that special ability to create blind loyalty and suspension of critical thinking, perfect for premium subscription retention.
Professional Poker Player: Masters of calculated risk, strategic deception, and reading psychological tells—ideal for pricing strategy.
Theme Park Architect: Experts at managing crowd flow, creating illusions of choice, and making people happily wait in lines for mediocre experiences.
Behavioral Economist: For nudging users into decisions they'll later regret but won't be able to pinpoint exactly why.
TikTok Content Strategist: Specialists in creating patterns so addictive they can make teenagers stare at rectangles for 6 hours instead of experiencing actual human connection.
Luxury Brand Storyteller: Experts at convincing people that a $50 product is worth $500 if you attach the right mythology to it.
Casino Experience Designer: Masters of creating environments where time and money lose all meaning until it's far too late.
Megachurch Pastor: For that special blend of community, FOMO, and soft extortion that keeps people coming back and opening their wallets.
Great post, and appreciate the comedic side of it all! I pray that at least some of it came from you, and it isn’t all entirely AI!
This is good read!! Fine, you convinced me to subscribe.. ;)