ChatGPT 101: The No BS Guide to Actually Getting Value From AI—Real Prompts, Real Situations

ChatGPT 101 is in session.

I looked everywhere for a practical ChatGPT guide to point people to - something with actual working prompts, not theory.

Couldn't find one that was both useful and detailed enough, so I built it myself.

Look, I hear this frustration every day. You know AI could be a massive helper for your work - comparing financial reports, analyzing deals, preparing documents - but when you try it, you get generic garbage that sounds like it was written by a robot who's never worked a day in their life. Your team is using it like Google with extra steps, typing in questions and hoping for magic.

Meanwhile, everyone claims they're "using AI" but you can tell they're just as lost as you are.

Here's what's actually happening: ChatGPT forgets your context mid-conversation, forcing you to repeat company details over and over. The output sounds so generic and untrustworthy that rewriting it takes longer than doing the work yourself. You need real solutions for real work - not theory about "the future of productivity."

You need it to sound like a human wrote it, with your tone and your specifics, not formal robot language that would embarrass you in front of clients. And most importantly, you need confidence it won't hallucinate numbers or make errors that destroy your credibility.

That's why I created this ChatGPT 101 course.

It's built from actual business frustrations, not marketing claims. These 36 specific prompts are organized into six categories that solve the exact problems you're hitting:

• Setup & Context - How to frame requests so AI actually remembers your situation (no more repeating yourself)
• Getting Better Output - Options before drafts, fix specific things, match your actual voice (not robot tone)
• Templates for Common Tasks - Client-ready emails, meeting summaries, decisions that save time not waste it
• Fix-It Tools - Emergency fixes when output is generic, wrong tone, or needs complete rewriting
• Organization & Processing - Batch similar work, clean sensitive data, stop babysitting the tool
• Verification & Safety - Catch hallucinations, verify numbers, know when to write it yourself

Every technique works in 60 seconds or less - instant, copy-paste solutions you can use on real work immediately. No theory, no "it depends," no becoming an AI expert. Plus I demo a bunch of them in the video and explain why they work!

Your financial report comparisons will have actual insights. Your emails will sound like you wrote them. Your meeting notes will turn into clear actions without 20 minutes of editing.

This is about making the tool trustworthy and useful for the work you're already doing, today, without wasting more time than doing it yourself.

Subscribers get all these newsletters!

Subscribed

How to Make AI Actually Useful in 60 Seconds or Less

==========================================================

Watch terrible, generic AI answers turn into something you'd actually send—in under a minute each.

HERE'S MY PROMISE (Yes, really) Every example in this guide works in 60 seconds or less. You'll copy it, paste it, and use what comes out. If something takes longer than making a sandwich, I cut it from this guide. No theory tours. No "it depends." Just working prompts you'll paste, run, and ship today.

Who needs this: People who tried ChatGPT three times and gave up because everything sounded like a robot wrote it. People stuck in generic-output hell who keep editing for 20 minutes. Anyone who keeps hearing "AI will change everything" but can't get it to write a decent email that sounds like them.

How to use this guide: Copy the 5-Minute Demo first (it’s in the video too)—actually run it right now if you can. Watch exactly what happens when you give AI structure instead of hope. Then find your specific situation in the Real Examples section. Keep the Fix-It Box handy for when things go sideways (they will, and that's fine).

Why Most People Give Up on AI (And Why This Time Will Be Different)

Let me tell you exactly what happened the last time many people that write me tried to use AI, because it happens to far too many people.

First, you typed something like "help me write an email to my boss" and got back three paragraphs that sounded like a robot doing an impression of a LinkedIn influencer from 2015. It started with "I hope this email finds you well" and you immediately knew you'd have to rewrite the whole thing. You closed the tab thinking this was all hype.

Second, even when you got something that wasn't terrible, you had no framework to judge if it was actually good. Should you edit it? Start over? How would you even know? Everything felt equally mediocre—not bad enough to throw away, not good enough to use. You were stuck in editing purgatory, spending more time fixing AI's output than it would have taken to write from scratch.

Third, when you got a bad answer (and you definitely did), you assumed one of two things: either you're somehow "bad at AI" (you're not), or AI is overhyped (it's not, when used right). Either way, you gave up. The tool that was supposed to save you time became another thing you failed at.

Here's what actually went wrong: You were talking to AI like it's Google—throwing a few keywords at it and hoping for magic. But AI isn't a search engine. It's more like a brilliant intern who knows everything but needs incredibly specific instructions to be useful.

Today you're getting four simple moves that fix this completely. You'll learn to ask for options before asking for a draft (this alone fixes 50% of problems). You'll tell it exactly what format you want back (no more surprise essays when you needed bullets). You'll fix specific things instead of starting over (turning garbage into gold with one more prompt). And you'll double-check anything important (because confident-sounding wrong answers have killed more projects than honest uncertainty ever did).

My promise: Your first actually useful result in 5 minutes. Not "maybe useful with editing." Not "I can see the potential if I squint." Something you'll actually send to a real person or publish with your name on it. Today.

Watch Me Do This: The 5-Minute Demo

This is where everything clicks. Copy this exactly—don't change a word yet. Run it in ChatGPT, Claude, or any AI chat. Watch what happens when you give clear, structured instructions instead of vague requests and hope.

The situation we're solving: You need to tell an important client you can't take their rush project, but you want to keep the relationship strong so they come back later. This is the kind of email that usually takes 30 minutes of agonizing over every word. We're going to nail it in 5 minutes.

Step 1: Set up the situation

Instead of just saying "help me decline a client project," you're going to paint the full picture. The AI needs context to give you something useful. Copy and paste this exactly:

Here's my situation: I run a design agency. A client who's been with us for 3 years wants a rush project.
What I need: Turn them down nicely without ruining the relationship.
My constraints: Keep it to 120-140 words. Can't use "bandwidth" or "capacity" (they're tired of hearing it).
Success looks like: They understand we can't help now but will definitely reach out again next quarter.

What just happened: You gave the AI four critical pieces of information it never would have guessed. It knows the stakes (3-year relationship), the constraints (word count and forbidden phrases), and most importantly, what success actually looks like. This is already 10x more context than most people ever provide.

Step 2: Be specific about who's reading

Now we're going to do something most people never think of—tell the AI exactly who will read this. Add this to your prompt:

Who's reading this: Their CMO, someone I've worked with for years
The tone I want: Friendly but professional, like talking to a colleague
Must include: Someone else they could work with and when we're free again

Why this matters: "Professional" means 50 different things. "Professional for a CMO I've worked with for years" means exactly one thing. The AI now knows this isn't a cold formal letter—it's a message between people with history.

Step 3: Get options before getting a draft

Here's the move that changes everything. Instead of asking the AI to just write something and hope for the best, you're going to make it think first. Add:

Before writing anything, give me 2 different ways I could handle this decline. Also ask me one question about the situation that would help you write better.

The AI will now give you two strategic approaches—maybe one focuses on recommending an alternative provider, while another emphasizes scheduling future work together. It'll also ask something insightful like "Why specifically are you declining?" or "What type of work would you accept from them?"

This is huge. The AI is now thinking about your problem, not just generating text. It's becoming a strategic partner, not a typing machine.

Step 4: Answer and pick your approach

Let's say the AI suggested:

• Approach A: Focus on referring them to a trusted partner
• Approach B: Emphasize your excitement for future collaboration

And asked: "What's the specific reason you're declining?"

You'll respond:

Let's go with approach B about future collaboration. We're declining because we're launching our own product and need everyone on deck through October. We can take new client work starting November 1st.

Notice what you just did: You gave the AI information it never would have guessed (product launch) and a specific date (November 1st). These details make the difference between generic fluff and an email that actually sounds real.

Step 5: Tell it exactly what format you want

Now we lock down the structure so you don't get a surprise novel when you wanted a note:

Now write the email with these exact parts:
- A subject line
- The email itself (120-140 words)
- Two backup subject lines I could use instead
Skip corporate buzzwords. Only apologize once at the start, then move on.

Why this works: You've eliminated all ambiguity. The AI knows exactly what sections to create, how long each should be, and what tone to avoid. No more getting a formal letter when you wanted a quick note.

Step 6: Fix specific things instead of starting over

This is where good becomes great. Instead of accepting the first draft or starting over (what most people do), you're going to make surgical improvements:

Look at what you wrote and tell me:
1. Is it specific enough? (Are we naming actual dates and next steps?)
2. Does it sound warm enough for a 3-year relationship?
List exactly 5 small things you'd change to improve it. Then rewrite it with ONLY those 5 changes. Show me a list of what you changed.

What's happening here: The AI will identify things like "Changed 'in the future' to 'November 1st'" or "Added personal touch about their last project." You get to see its thinking, and the revision only changes what needs changing—the rest stays stable.

Step 7: Double-check before sending

Last step—make sure this won't backfire:

For this final version, tell me:
- What assumptions are you making?
- How confident are you (0-100%) this keeps the relationship strong?
- What's one thing I should double-check before hitting send?

The AI might say it's assuming they'll appreciate the referral, it's 85% confident this maintains goodwill, and you should double-check that the partner you're recommending has capacity. This kind of reflection catches problems before they happen.

What you just got: An email that sounds like you wrote it, preserves a valuable relationship, includes specific dates and next steps, and took less time than reading articles about how to write emails. Compare this to your first "help me write an email" attempt. The difference isn't the AI—it's the process.

The 12 Moves That Actually Work (With Full Explanations)

Each move I'm about to show you has three parts: why it works (the psychology and mechanics behind it), how to actually do it (specific steps), and something you can copy-paste right now to try it yourself. Start with the Basic Four—master those before moving on. Everything else builds on that foundation.

THE BASIC FOUR (Learn These Today)

1. Start by explaining the situation

Why this works: Think about giving directions to your house. Would you just say "help me get home" or would you say where you're starting from, where you're going, and what landmarks to watch for? AI is the same. It can generate infinite possibilities, but without context, it's just guessing. When you explain your starting point, destination, constraints, and success criteria, you narrow infinity down to exactly what you need.

Most people treat AI like a magic oracle that should somehow know their situation. But AI has no idea if you're a CEO or a student, if this is urgent or whenever, if the stakes are high or low. By spending 30 seconds setting context, you eliminate 90% of useless outputs.

How to do it: Before asking for anything, write four simple lines. Think of it as briefing a smart colleague who just walked into the room. They need to know: What's the situation? What are we trying to achieve? What can't we do? How will we know we succeeded?

Try this:

Here's my situation: [What's happening right now - one sentence]
What I need: [What would success look like - be specific]
My constraints: [What I can't do, say, spend, or use]
Success looks like: [How I'll know this worked - measurable outcome]

Real example that worked: Sarah was struggling with a process change proposal. She kept getting generic business speak from AI. Then she added this line: "Success looks like: My manager agrees to try the new process for 2 weeks as a pilot." That single line changed everything. Instead of vague suggestions about "driving alignment" and "leveraging synergies," she got specific language about pilot programs, success metrics, and rollback plans. The proposal was approved in one meeting instead of three rounds of revisions.

2. Get options before drafts

Why this works: Imagine you're at a restaurant and the waiter just brings you food without asking what you want. That's what happens when you ask AI to write something immediately. It picks a direction and runs with it, and that direction might be completely wrong for your situation.

When you force AI to show you different approaches first, two things happen. First, you get to pick the strategy that actually fits your situation. Second, the AI has to think about your problem before writing, which means it understands the nuances better. It's the difference between a thoughtful recommendation and a lucky guess.

How to do it: Add a speed bump between request and output. Make the AI slow down, consider options, and ask clarifying questions. This feels like it takes longer, but it actually saves time because you don't waste cycles on wrong approaches.

Try this:

Before writing anything, give me 2 different approaches to this. Also ask me 2 questions that would help you do this better. I'll answer, then you can write.

What happens next: The AI will outline two strategies—maybe one conservative, one bold. It'll ask something like "What's your relationship with this person?" or "Have you tried addressing this before?" Your answers become crucial details that make the final output actually useful instead of generically acceptable.

3. Be specific about your audience

Why this works: "Professional" is meaningless. Professional for a Wall Street banker is different from professional for a startup founder, which is different from professional for an academic. When you tell AI "write professionally," you're basically saying "guess what I mean and hope for the best."

But when you say "write for my boss who hates long emails and loves data," suddenly the AI knows exactly what to do. Short paragraphs. Bullet points with numbers. No fluff. One clear ask. The specificity transforms generic into perfect.

How to do it: Before any writing task, picture the actual person reading it. What do they care about? How do they like information presented? What triggers them? Share these details with the AI like you're briefing a ghostwriter.

Try this:

Who's reading: [Specific person and their role - include personality hints]
Tone: [How it should feel - use comparisons like "like talking to..."]
Length: [Exact word count or time to read]
Don't include: [What this person hates or what would fail]

Real-world win: Marcus needed to email his company's board about a missed target. First attempt with "write professionally" got him corporate word salad. Second attempt with "Who's reading: Board members who want brutal honesty and solutions, not excuses" got him an email that led to productive discussion instead of finger-pointing. The board chair actually complimented his clarity.

4. Tell it exactly what format you want

Why this works: You know that frustration when you ask for a quick summary and get a five-paragraph essay? Or when you need detailed analysis and get three bullet points? That's because AI defaults to whatever format it feels like unless you specify exactly what you want.

This is like the difference between telling a designer "make it nice" versus giving them a wireframe. When you specify exact sections, word limits, and what to exclude, you eliminate format surprises. You get what you expected, which means you can actually use it.

How to do it: Think like you're creating a template. List every section you want, in order, with constraints for each. Be explicit about what you don't want—AI loves to add "helpful" extras that just create work for you to delete.

Try this:

Give me exactly these sections:
- Summary (5 bullet points, one line each, facts only)
- Options (3 choices with pros and cons, 2 of each)
- Risks (Low/Medium/High with specific scenarios)
- Next Steps (5 actions with who does them and when)
Keep each section under 200 words. Don't explain what these sections are, just fill them in.

Why the last line matters: That instruction about not explaining sections? That alone saves you from paragraphs of "The summary section provides an overview of..." that you'd have to delete. Small instructions, big time savings.

Put it together into a catchall prompt you can take with you:

I have a situation not covered in the guide. Here's my context:
- Situation: [Describe your scenario in one sentence]
- Goal: [What outcome you need]
- Constraints: [Word count, tone, must-avoid/must-include]
- Audience: [Who will read it and how they think]
Using the guide’s methods (Ask→Do, clear format, delta-only revisions, verification), produce:
1. A brief outline of 2 approaches
2. The full prompt to paste into AI—including context, ask→do instructions, format specification, and verification step
3. A “why this works” note  

MOVES THAT MAKE IT BETTER (Learn These Next)

5. Fix specific things instead of starting over

Why this works: Here's what most people do: they get an AI output that's 70% good, and either accept mediocrity or start completely over. Both are mistakes. Starting over is like shuffling a deck of cards—you might get better cards, but you might get worse ones, and you've lost what was working.

But when you identify specific problems and fix only those, you keep what works and improve what doesn't. It's like editing a photo—you don't take a new picture, you adjust the contrast. Version 2 that changes everything is just a different document. Version 2 that fixes five specific things is actually better.

How to do it: Train yourself to spot specific problems, not vague dissatisfaction. Instead of "this doesn't feel right," identify "the opening is too weak" or "the data section needs examples." Then tell the AI exactly what to fix and—crucially—what to leave alone.

Try this:

Look at this and tell me what could be better. List exactly 5 specific changes. Then rewrite it making ONLY those 5 changes. Show me a list of what you changed.

What you'll see: The AI will list things like "Changed 'in the near future' to 'by March 15'" and "Added specific example in paragraph 2." You can see its thinking, agree or disagree with changes, and the document evolves instead of randomly regenerating.

6. Make it sound like you

Why this works: Every person has verbal fingerprints—words they use, rhythms they follow, levels of formality they're comfortable with. When you tell AI "write like me," it has no idea what that means. Are you formal? Casual? Do you use contractions? Long sentences or short? It's guessing blind.

But when you give it an actual sample of your writing, it can detect patterns. It notices you start sentences with "Look," or that you use lists of three, or that you explain with examples. It's like the difference between describing someone's face versus showing a photo.

How to do it: Find something you wrote that people responded well to. An email someone said was "perfectly clear." A document your boss praised. A message that got the result you wanted. Even one paragraph is enough. This becomes your style fingerprint.

Try this:

Make this sound like my writing style. Here's an example of how I write:
"[paste a paragraph you wrote that worked well]"
Keep the facts the same but match this tone and rhythm.
Pro tip: If you can't find an example, here's a cheat code that works for most people:
Make it shorter and plainer. Use simple words. Remove all business jargon. Write like you're explaining to a smart friend who doesn't work here.

7. Use templates for messy information

Why this works: Your brain doesn't process information soup very well. That's why meeting notes that are just stream-of-consciousness typing are useless a week later. You can't find what you need, can't remember what was decided, can't track what should happen next.

Templates force structure onto chaos. They're like putting items into labeled boxes instead of dumping everything in a pile. When AI organizes information into consistent sections, your brain can process it faster, you can find what you need, and nothing important gets lost.

How to do it: Before dumping information into AI, decide what boxes you need. Think about what you'll want to find later. What questions will people ask? What will you need to track? Create sections for those specific needs.

Try this:

Organize this using these exact sections:
Summary: (5 main points - what actually happened)
Decisions Made: (who decided what, when)
Open Questions: (what needs answers, who's answering)
Risks: (what could go wrong, how likely)
Next Week: (who's doing what by when)

Hidden benefit: Once you have a template that works, you can use it forever. Same sections every time means your brain builds patterns. You'll process information faster because you know where to look.

8. Remove private information first

Why this works: That horror story about Samsung employees accidentally sharing trade secrets with ChatGPT? That's what happens when you paste first and think later. Client revenue, employee names, deal details—none of this belongs in AI systems that might use it for training or store it who-knows-where.

But you still need AI's help with sensitive documents. The solution is surgical redaction—replace specifics with placeholders that preserve meaning but remove identifying details. The AI can still help with structure, tone, and strategy without ever seeing the actual sensitive data.

How to do it: Before pasting anything, scan for names, numbers, dates, and companies. Replace them with generic markers that preserve the relationship but hide the identity. "John from Sales" becomes "[Sales Manager]". "$2.3M deal" becomes "[$2-3M deal]". "March 15 deadline" becomes "[Q1 deadline]".

Try this:

First, clean up this text for privacy:
- Replace names with [Job Title]
- Replace exact dollars with ranges
- Replace specific dates with quarters
- Replace company names with [Industry Type + Size]
Just show me the cleaned version.

What people miss: It's not just about protecting secrets. It's about being able to use AI for sensitive work at all. Once you build this habit, you can get help with performance reviews, salary negotiations, and strategic planning without risk.

9. Double-check important stuff

Why this works: AI can sound absolutely certain while being completely wrong. It's not lying—it's doing sophisticated pattern matching without true understanding. The same confidence that makes it convincing makes it dangerous for facts, figures, and anything with real consequences.

But when you force AI to show its work and rate its confidence, something interesting happens. It often catches its own errors. It'll say "I'm assuming X but you should verify Y." This self-reflection is like having a smart colleague say "I think this is right but you should double-check with legal."

How to do it: For anything with consequences—numbers, claims, recommendations—add a verification step. Make the AI explain its reasoning, state assumptions, and tell you exactly how to fact-check. This takes 30 seconds and prevents disasters.

Try this:

After your answer, add:
- What you're assuming (bullet list)
- How confident you are (0-100%)
- How I could fact-check the main point in 5 minutes

Real save: Jennifer was using AI to help with investor materials. The AI confidently stated a market size figure. She asked for verification steps, followed them, and discovered the number was off by 10x. The 30 seconds of verification saved her from a credibility-destroying mistake in front of investors.

ADVANCED MOVES (Try These After a Week)

10. Process multiple things at once

Why this works: Switching contexts is expensive for your brain. Opening AI, explaining your situation, getting an answer, closing it, then doing it again for something similar—that's like restarting your computer between every task. Batching similar work eliminates that overhead.

But the real power is consistency. When you process three emails together, they naturally align in tone and style. When you summarize three meetings at once, you spot patterns you'd miss doing them separately. It's the difference between cooking three separate meals and meal-prepping for the week.

How to do it: Group similar tasks—all your email responses, all your meeting summaries, all your content ideas. Use clear separators so outputs don't blur together. Include full context for each item (the AI doesn't carry context between them unless you repeat it).

Try this:

I have 3 things to process. Handle them separately and give me 3 responses back.
First thing:
[complete context and instructions]
====
Second thing:
[complete context and instructions]
====
Third thing:
[complete context and instructions]

Common mistake to avoid: People try to batch different types of tasks and wonder why quality drops. Batch similar things—three client emails, not an email plus a report plus a slide deck. The similarity is what creates efficiency.

11. Think through big decisions properly

Why this works: Most decision-making with AI is terrible. People ask "should I do A or B?" and get a pros and cons list that tells them nothing new. That's because they're asking the wrong question. The right question isn't "which is better?" but "what would have to be true for this to fail?"

When you force analysis of failure modes, hidden assumptions, and trigger points, you get actual insight. It's the difference between a weather forecast and understanding why storms form. One helps you decide whether to bring an umbrella. The other helps you plan a sailing trip.

How to do it: Structure your decision analysis like a strategic consultant would. What factors matter most? How does each option score? What would change your mind? What are you not considering? This framework turns AI from a Magic 8-Ball into a strategic thought partner.

Try this:

Help me decide between [Option A] and [Option B]:
1. What matters most in this decision? (weight factors)
2. Score each option (0-10) with reasoning
3. What would have to change to flip this choice?
4. How could the winner fail catastrophically?
5. What warning signs should I watch for in week 1?
6. What's the specific plan for the first 7 days?

What this reveals: You'll often discover that what seemed like a close decision isn't. Or that you're optimizing for the wrong thing. Or that there's a Option C you hadn't considered. The structured analysis surfaces insights that free-form discussion misses.

12. Know when to write it yourself

Why this works: Some things are too nuanced, too sensitive, or too important for AI to handle alone. A termination letter. A personal apology. A make-or-break pitch. These need human judgment, emotional intelligence, and accountability that AI can't provide.

But AI can still help without doing the actual writing. It can suggest openings when you're stuck. It can provide language for difficult passages. It can spot issues you might miss. Think of it as a writing coach rather than a ghostwriter for sensitive content.

How to do it: For sensitive content, explicitly tell AI not to write the final version. Ask for components, options, and guidance instead. You maintain control and responsibility while still getting help with the hard parts.

Try this:

Don't write this for me. Instead:
- Give me 3 different ways to open
- Suggest specific language for the difficult part
- Tell me what I should verify independently
- Flag anything that seems too risky to automate
I'll assemble the final version myself.

When this matters most: Legal communications, personnel issues, anything involving strong emotions, public statements, and anything where being wrong has serious consequences. The AI helps you think, but you own the words.

Real Examples: From Terrible to Useful

Here are eight actual scenarios people face every day. For each one, I'll show you the lazy prompt that everyone tries first (and fails), then the structured prompt that actually works, and how to tell if it succeeded. Find the one closest to your situation and copy it exactly.

1. Difficult Conversation with Your Boss

This is the conversation you've been avoiding. Maybe it's about workload, or boundaries, or that thing they do that drives everyone crazy. Most people either say nothing or blow up. Here's how to prepare properly.

What most people type:

Help me tell my manager to stop changing deadlines.

This gets you generic advice about "communication" and "alignment" that sounds like it came from a corporate training video no one watches.

What actually works:

Situation: My manager changes project deadlines weekly, usually mentions it in passing during other conversations
What I need: Set up a simple process for deadline changes without seeming difficult or rigid
Can't do: Sound accusatory, damage our good relationship, or seem like I can't handle change
Success: We agree on a system for handling future deadline changes that works for both of us
Who's reading: My manager who values flexibility and speed over process
Tone: Collaborative problem-solving, not complaining
Get options first: Give me 3 ways to open this conversation and ask what else would help
Format I want:
- Opening (acknowledge something positive about their style)
- The issue (impact on team without blame)
- Suggested solution (specific process that's lightweight)
- Example of it working elsewhere
- Close (ask for their input and improvements)
Also add: If they push back saying process slows things down, I could say [what].
If they agree, next step is [what].
Double-check: List your assumptions about their motivations and tell me if this will likely go well

What changed: You went from venting about a problem to designing a solution. You anticipated objections. You made it about improving work, not criticizing behavior. The AI knows this is about finding common ground, not winning an argument.

How you know it worked: You actually have the conversation within a week (instead of stewing for another month). They agree to try something, even if modified. The chaos decreases by at least half. Team stress goes down. You stop dreading Monday morning timeline changes.

2. Declining Something Politely (But Firmly)

Saying no without burning bridges is an art. Too harsh and you damage relationships. Too soft and they don't hear the no. Here's how to hit the sweet spot.

What most people type:

Write a professional rejection email.

This gets you a cold, formal template that sounds like a legal document and leaves everyone feeling bad.

What actually works:

Situation: Turning down a speaking opportunity at a conference I've done before
Who's getting this: Conference organizer I've worked with twice, knows me well
Tone: Warm but clear, peer-to-peer, not corporate to vendor
Length: 100-120 words max
Must include:
- Real reason (our Q2 product launch needs all hands)
- Someone else they could ask (with specific reason why they're good)
- That I'm interested in next year (with specific detail)
Must NOT include: Over-apologizing, generic "bandwidth" excuse, leaving door open for negotiation
Format: Subject line + Email + 2 alternative subject lines
Then: Tell me 5 small ways to make it warmer, rewrite with only those changes
Show me exactly what you changed and why

What changed: Specific details (who's reading, real reason, alternative solution) transformed corporate rejection into helpful redirection. The organizer feels supported, not rejected.

How you know it worked: They reply positively, often thanking you for the referral. Future opportunity gets mentioned specifically. The whole exchange takes 5 minutes instead of 30 minutes of agonizing. You feel helpful instead of guilty.

3. Understanding Something Complex

We all hit topics that make our brains hurt. Technical concepts, financial instruments, legal frameworks. Here's how to get explanations that actually stick.

What most people type:

Explain options trading like I'm five.

This gets you either patronizing oversimplification or still-confusing jargon, with no middle ground.

What actually works:

Explain [options trading] for someone who understands [how stocks work, basic supply/demand] but nothing about [derivatives, options specific terminology]
Give me:
- The core idea in 5 simple points (one sentence each)
- One real-world comparison (not stock-related)
- An actual example with real numbers and dates
- 3 things I need to understand first (prerequisites)
- 3 questions to test if I really get it (increasing difficulty)
- Your confidence level that this explanation is accurate (0-100%)
- One common misconception people have
If I still don't get it, what's probably confusing me?

What changed: Starting from what you DO know (stocks) instead of assuming zero knowledge. Adding self-tests to verify understanding. Asking for prerequisites helps you identify gaps.

How you know it worked: You can answer all three test questions without looking back. You could explain it to someone else using your own words. The concept clicks and stays clicked. You spot references to it in articles and actually understand them.

4. Turning Meeting Notes into Actions

Every meeting generates information soup. Here's how to turn three pages of notes into clear actions that actually happen.

What most people type:

Summarize my meeting notes.

This gets you a shorter version of the same mess, with no clarity on who does what by when.

What actually works:

[Paste your messy notes here]
Pull out and organize:
- Summary: 5 main things that actually happened (not topics discussed)
- Decisions: Who decided what, when (name names)
- Open Questions: What needs answers (question → who's answering → by when)
- Risks: What could go wrong (risk → likelihood → impact → who's watching it)
- Next 7 Days: Specific actions (action → owner → deadline → success measure)
Skip any section with nothing in it. Don't add fluff or context.
Flag any action item that doesn't have a clear owner or deadline.
If there's conflict between different parts of the notes, flag it.

What changed: Template forces accountability (names and dates). Risks section catches problems before they happen. Success measures prevent "I thought you meant..." confusion.

How you know it worked: Zero "wait, what did we decide?" messages. 80% of actions complete on time. Follow-up meetings are shorter because everyone knows status. You can find what you need in 10 seconds, not 10 minutes.

5. Creating Content Ideas That Don't Suck

Everyone needs content ideas. Most are generic garbage. Here's how to get ideas that actually resonate with your specific audience.

What most people type:

Give me blog post ideas about remote work.

This gets you "10 Tips for Working from Home" and other titles that have been written 10,000 times.

What actually works:

Topic: Remote work productivity
Who reads this: Managers with 5-15 person teams, first time managing remotely, tech industry
My voice: Practical, been-there-done-that, slightly skeptical of "best practices," no fluff
Goal: Position myself as the realistic remote team expert, not another "future of work" theorist
Create:
1. One detailed article outline (8 sections with 2 specific points each, include data/examples)
2. 5 short social media posts (3 sentences each, different angles, specific not generic)
3. 10 attention-grabbing headlines (under 10 words, specific not vague)
4. For each short post: one specific example or story so it's not generic
5. 3 contrarian takes that go against common remote work advice
Check: Flag any corporate clichés and suggest replacements
Tell me which idea is most likely to get shared and why

What changed: Defining exact audience (managers, not employees) and voice (skeptical, not cheerleader) killed the generic. Every idea targets a specific person with a specific problem.

How you know it worked: You publish 3 pieces this week instead of staring at blank page. Engagement goes up 2x because content resonates. One piece becomes your most-shared ever. People DM you saying "this is exactly what I needed."

6. Making a Decision (With Actual Analysis)

Big decisions paralyze people. AI usually makes it worse with endless pro/con lists. Here's how to get real clarity.

What most people type:

Should I use Notion or Monday.com?

This gets you feature comparisons you could have googled yourself.

What actually works:

Decision: Notion vs Monday.com for my 20-person product team
Context: Team is technical but resistant to new tools, we use Slack/GitHub/Figma currently
Consider these factors (you rate importance 1-10):
- How fast the team can learn it
- Integration with our current tools
- Total cost for 2 years
- Features specific to product teams
- Mobile experience quality
- Vendor lock-in risk
Score each option (0-10) on each factor and explain why
Show me the weighted scoring
What would have to be true to flip this decision?
Top 3 ways the winner could fail for our specific situation
Warning signs to watch for in the first 2 weeks
Week 1 implementation plan with specific milestones
What am I not considering that could matter?
What changed: Forcing weighted criteria reveals what actually matters. Failure analysis prevents expensive surprises. Week 1 plan makes decision real, not theoretical.

How you know it worked: Decision made in one sitting instead of weeks of waffling. No buyer's remorse in week 2. Team actually adopts the tool. You can explain the decision rationale months later.

7. Fixing Technical Problems

Tech problems are frustrating because error messages are useless and Google gives you 47 different solutions. Here's how to get targeted help.

What most people type:

Excel shows error help

This gets you generic troubleshooting that assumes you know nothing or everything, with no middle ground.

What actually works:

My setup: Excel 365 on Windows 11, using VLOOKUP formula
Problem: Getting #VALUE! error in cells B2 through B50 when the formula worked yesterday
Formula I'm using: =VLOOKUP(A2,Sheet2!A:C,3,FALSE)
Already tried: Checked cell formats, looked for extra spaces, verified sheet names
Tell me:
1. Three most likely causes for this specific situation (most probable first)
2. Quick way to check each one (under 30 seconds each)
3. Step-by-step fix for the most likely cause
4. If that doesn't work, what diagnostic question would narrow it down?
5. How to undo changes if I make it worse
6. Plain English explanation of what probably went wrong
Be specific: Tell me exactly which menu and button for each step (don't assume I know)
Include keyboard shortcuts where relevant

What changed: Context about what worked before, specific formula, and what you already tried eliminates generic advice. Undo instructions prevent making it worse.

How you know it worked: Fixed in under 10 minutes without calling IT. You understand why it broke so it doesn't happen again. You can help others with same problem.

8. Learning Something New (That Actually Sticks)

Most learning with AI is random—"teach me Python" gets you Chapter 1 of a textbook. Here's how to build real skills.

What most people type:

Teach me Python

This gets you either "Hello World" basics you'll forget or advanced concepts you're not ready for.

What actually works:

Current skill: I can use Excel formulas including VLOOKUP and pivot tables, never coded before
Goal: Automate my weekly sales report that takes 2 hours in Excel
Timeline: 30 minutes a day for 7 days (total 3.5 hours investment)
Create:
- Day 1-7 plan (one concept + hands-on practice per day)
- Each day builds on previous (no jumping around)
- Day 4 should review and reinforce days 1-3
- Day 7: Small project I can actually finish and use
- 5 key concepts as simple flashcards
- 3 common beginner mistakes and how to avoid them
- Exact tools to install (with links)
Use my actual work context: sales data, CSV files, weekly reports
Skip anything I won't need for my specific goal
Include: Free website where I can practice without installing anything
What's the minimum I need to learn to achieve my specific goal?

What changed: Specific goal (automate one report) instead of vague "learn Python." Real timeline creates urgency. Using actual work context makes it immediately relevant.

How you know it worked: You complete the day 7 project and it actually works. Your weekly report now takes 10 minutes instead of 2 hours. You can modify the code for similar tasks. You're googling specific Python questions, not "how to start with Python."

The Fix-It Box (When Things Go Wrong)

Keep this section bookmarked. When AI gives you garbage output, don't start over and don't give up. One of these fixes will solve 90% of problems in 30 seconds.

Problem: Sounds like a robot wrote it

The fix:

You missed what I need. The tone is too formal/corporate. Make this sound like a human talking to another human. Short sentences. Simple words. Like explaining to a smart colleague over coffee. Remove all business jargon and corporate speak.

Why this works: "Human" is too vague, but "colleague over coffee" is specific. Short sentences force clarity. Banning jargon kills the robot voice.

Problem: It completely missed the point

The fix:

You misunderstood what I'm trying to achieve. My ACTUAL goal is [specific outcome]. The real constraint is [specific limitation]. Forget everything else and focus only on [core need]. Try again with just that focus.

Why this works: Clear reset with single focus. Sometimes AI overcomplicates because you gave it too much context.

Problem: The tone is all wrong

The fix:

This tone is completely wrong for my situation. Here's an example of the right tone: "[paste a paragraph with the right feel]". Match that energy and style exactly. Keep the facts but change how it feels.

Why this works: Showing beats telling. The example gives AI a target to match instead of guessing what you mean.

Problem: Way too long and rambling

The fix:

Cut this to exactly [X] words. Lead with the main point in the first sentence. One idea per sentence. Remove: all setup, throat-clearing, context-setting, and repetition. Make every word earn its place.

Why this works: Specific word count forces decisions. "Lead with main point" fixes buried ledes. "One idea per sentence" creates clarity.

Problem: Too generic and vague

The fix:

This is too generic. Add: Real numbers, exact dates, specific examples, actual names (or role titles), concrete next steps. Remove: "various," "multiple," "several," "etc," "and so on." Every point needs specifics.

Why this works: Banning weasel words forces precision. Demanding examples makes it real.

Problem: Sounds confident but might be wrong

The fix:

Show me your reasoning for the main claim. What sources or logic are you using? Rate your confidence 0-100%. What could be wrong here? How would I verify this independently in 5 minutes?

Why this works: Forces AI to show its work. Often reveals assumptions or uncertainty it was hiding.

Problem: Too many options, need decision

The fix:

Too many choices. Given my specific situation [remind key context], pick the ONE best approach. Explain why it's best for me specifically. Give me just that one, polished and ready to use. No alternatives.

Why this works: Analysis paralysis is real. Sometimes you need AI to make the call.

Problem: Added complications that don't exist

The fix:

You're overcomplicating this. Assume everything is normal and standard unless I specifically said otherwise. Don't invent edge cases or unusual scenarios. Give me the straightforward solution for the typical situation.

Why this works: AI sometimes invents problems to sound thorough. This forces simplicity.

Ready-to-Copy Templates for Common Situations

These are complete templates for situations you'll face this week. Don't modify them yet—use them exactly as written first, then adjust once you see how they work.

Template: Difficult Conversation Prep

I need to have a difficult conversation
Situation: [What's happening - include history]
My goal: [What I want to achieve]
Their likely position: [What they probably want]
Relationship to preserve: [Why this matters]
Give me:
- 3 different ways to start this conversation (different tones)
- Main points as a brief list (no more than 5)
- If they say [likely objection], I should say [what]
- If they get defensive, I should [what]
- If they agree, the immediate next step is [what]
- Language to repair if it goes badly
- One thing definitely NOT to say and why
Check: What are you assuming about their motivations?
How likely is this to preserve the relationship (0-100%)?
What's the worst realistic outcome and how do I prevent it?

Template: Make This Writing Better

Here's what I wrote: [paste it]
Who's reading: [specific person/role and what they care about]
What should happen after they read it: [specific action/feeling]
Main problem with current version: [too long/too corporate/too vague/wrong tone]
Check for:
- Vague phrases that need specifics (list them)
- Corporate speak to remove (list it)
- Missing examples or evidence (what needs support)
- Unclear call-to-action (what should they do next?
List 5 specific improvements with reasoning.
Rewrite with ONLY those changes.
Show before/after for each change.
Rate the improved version (0-100%) for achieving my goal.

Template: Explain Like I'm Smart But New

Explain [complex topic]
I already understand: [related concepts I know]
I'm confused about: [specific part that's blocking me]
Why I need to know this: [practical application]
Give me:
- 5-point summary in plain English (no jargon)
- One comparison to [something familiar to me]
- Concrete example with real numbers/scenario
- 3 prerequisites I might be missing (with 1-sentence explanations)
- 3 test questions (easy → medium → hard)
- Most common misconception and why it's wrong
- Where to learn more (specific resource not generic advice)
How confident are you this explanation is accurate? ____%
If I'm still confused after this, what's probably the issue?

Template: Email That Gets a Response

Situation: [Why I'm writing]
Recipient: [Role, how well I know them, how busy they are]
My goal: [What I need from them]
Their benefit: [What's in it for them]
Constraints: [Maximum length, timeline]
Write email with:
- Subject line that gets opened (not generic)
- Opening that acknowledges their situation
- Clear ask in first paragraph
- Why this matters to them specifically
- Exactly what I need and by when
- Easy way to say yes
- Graceful way to say no
- P.S. line that reinforces urgency/importance
Check: Reading time under 45 seconds?
Clear action needed? Easy to forward?

Template: Turn Data Into Story

Data I have: [paste key numbers/findings]
Audience: [Who needs to understand this]
Their context: [What they know/don't know]
Decision needed: [What they should do with this info]
Their biggest concern: [What they're worried about]
Create:
- One-sentence headline (the "so what")
- 3-paragraph narrative (situation → insight → action)
- Supporting data woven in (not listed)
- One powerful comparison they'll remember
- Address the obvious objection
- Clear next step with deadline
Make the numbers meaningful: Compare to something they know
Highlight the surprising part
End with specific action not generic "let's discuss"

Making It Sound Like You (Your Voice Fingerprint)

"Professional tone" is where personality goes to die. Here's how to keep sounding like yourself, not a corporate bot.

The problem with tone instructions: When you say "professional but friendly," AI has no idea what that means. Professional for a law firm? Friendly like customer service? It's guessing blind, and usually guesses wrong.

Find your voice sample: Look through your sent emails for one that worked. Someone replied quickly. Your boss said "clear and to the point." A client said "this is why I work with you." That's your voice fingerprint.

Try this:

Make this match my writing style. Here's how I actually write:
"[paste your paragraph]"
Keep this exact voice. Notice I [specific pattern, like "use short sentences" or "start with the point"].
What if you can't find an example?

Here's a cheat code that can help:

Make it shorter and plainer. Use simple words a smart 12-year-old would understand. Remove ALL business jargon. Write like you're explaining to a friend who doesn't work here. Start with the point, not the context.

Voice patterns to preserve:

• If you use contractions (don't, won't), tell AI to use them
• If you start sentences with "Look," or "So," or "Here's the thing," mention it
• If you use lists of three, or rhetorical questions, or specific phrases, point them out
• If you're direct vs. indirect, specific vs. general, formal vs. casual—say so

What ruins your voice:

• Apologizing when you don't normally ("I hope this finds you well")
• Fancy words you'd never use ("leverage," "synergy," "align")
• Long sentences when you write short
• Explaining obvious things
• Corporate hedging ("perhaps we might consider")

Processing Multiple Things at Once (Batch Magic)

Stop having the same conversation with AI three times. Here's how to handle multiple similar tasks in one go.

Why batching helps: Your brain takes 15 minutes to fully focus on a task. When you handle three emails separately, that's 45 minutes of context-switching. Handle them together, it's 15 minutes total. Plus AI maintains consistency across all of them.

Try this:

I need to process 3 separate things. Handle each one individually but maintain consistent tone.
First item:
[Complete context and instructions]
[Include all constraints]
[Specify format needed]
====
Second item:
[Complete context and instructions]
[Include all constraints]
[Specify format needed]
====
Third item:
[Complete context and instructions]
[Include all constraints]
[Specify format needed]

Perfect for batching:

• Email responses to different people about same topic
• Meeting summaries from same day
• Performance reviews for team members
• Content variations for different platforms
• Status updates for multiple projects
• Feedback on multiple documents

What breaks batching:

• Mixing different types (email + report + analysis)
• Forgetting to repeat context for each item
• Not using clear separators
• Assuming AI remembers item 1 when working on item 3

Create a template for the repeated parts:

[TEMPLATE for all items]
Tone: Warm but professional
Length: 100-150 words
Format: Opening, main point, next step
Avoid: Apologizing, corporate speak
Item 1: Declining a meeting (context: ...)
====
Item 2: Declining a project (context: ...)
====
Item 3: Declining a committee role (context: ...)

What Can Go Wrong (And How to Fix It)

Let's be honest about what breaks and how to fix it, so you don't give up when you hit these common problems.

Problem: Everything sounds the same

• Cause: Using same template repeatedly without voice variation
• Fix: Add your voice sample to EVERY prompt
• Fix: Vary your opening instructions
• Fix: Ask for 3 versions and pick the most "you"

Problem: Good enough but not great

• Cause: Accepting first draft instead of iterating
• Fix: Always do the "5 specific improvements" step
• Fix: Use "make it 20% more [specific quality]"
• Fix: Compare to your best past work

Problem: Takes longer than writing myself

• Cause: Over-engineering simple tasks
• Fix: Simple task = simple prompt
• Fix: Save templates for repeated tasks
• Fix: Batch similar work together

Problem: Confidently wrong information

• Cause: AI pattern-matching without understanding
• Fix: Add verification step for ANY facts
• Fix: Check numbers twice
• Fix: When in doubt, mark as "VERIFY" for yourself

Problem: Inconsistent quality

• Cause: Vague instructions, different each time
• Fix: Use exact same template for same type of task
• Fix: Include successful example in prompt
• Fix: Start with working template, then modify

This approach doesn't work for:

• Really creative work → Get options and ideas, but you create
• Sensitive personnel issues → Get coaching on approach, write yourself
• Breaking news or live facts → Verify everything independently
• Legal or medical advice → Get professional help
• Anything where being wrong has serious consequences → Human judgment required

Your Permanent Copy-Paste Toolkit

Save these snippets in a document you can access quickly. Copy and paste as needed—they work for any type of request.

For any request - Structure and format

Give me exactly what I asked for in the format I specified. Maximum 200 words per section unless I said otherwise. Don't explain what these sections are or add context, just fill them in.

For any request - Make it better

Review this for [clarity/brevity/impact/warmth]. List exactly 5 specific improvements with reasoning. Rewrite implementing ONLY those 5 changes. Show me a before/after comparison of what changed.

For any request - Verify accuracy

After your response, add:
- Key assumptions you're making (bullet list)
- Confidence level: [0-100%] with reasoning
- How to verify the main claim in under 5 minutes
- What could be wrong or missing

For any request - Privacy protection

I should note this does not mean that you can use this step and avoid red data requirements at your work. But it can be helpful if you are operating within policy! Before processing, clean this for privacy:

Replace: names→[Role], companies→[Industry Size], dollars→[range], exact dates→[quarters], locations→[regions]
Return ONLY the cleaned version, then proceed with the task.

For any request - Match my style

Make this sound like I wrote it. Here's my style:
"[your saved voice sample]"
Match this tone, rhythm, and level of formality exactly.

For any request - When it's wrong

This isn't what I need. The core issue is [specific problem]. Forget everything except [key goal]. Give me the simplest solution that achieves [specific outcome]. No extra features or edge cases.

If you’re still reading down here, well done. I put a lot of heart into this. Think of it as Nate’s ChatGPT 101 course—all the most practical use-cases I can think of in one place.

I hope you find it helpful! Share it with those who need it.

I make this Substack thanks to readers like you! Learn about all my Substack tiers here

Subscribed