Introduction
Why AI matters
If you’ve tried ChatGPT or other AI tools and felt underwhelmed, you’re not alone. Most people get generic results and dismiss it as overhyped. But the issue isn’t the tool. It’s how you use it.
AI isn’t going anywhere. Just as computer literacy became essential in the 2000s, AI literacy is now a baseline skill. Even companies like Shopify expect it (see their CEO’s memo). Master it now, and you’ll work faster, do more with less, make better decisions, and stand out.
One caveat before we begin
AI isn’t magic. It won’t replace your judgment or do everything for you.
It has limits and can feel dumb at times. Knowing its strengths and weaknesses helps you use it effectively. Use well it can feel like you have a team of tireless employees and consultants.
What this is (and what it’s not)
This isn’t just about writing better prompts. It’s about using AI effectively in your work and life.
Treating AI like a vending machine leads to frustration. Instead, think of it as a teammate.
Good prompting is good delegation: break down goals, assign clear tasks, and give feedback.
This is how I use it daily.
📘 Prompting Basics
We’ll start with prompts because it’s the core piece of interacting with these AI models.
What’s a prompt?
Your prompt is what you type into ChatGPT. They’re instructions that tell the model what output you want.
- “Write me a casual tweet about this article”
- “Turn this into a 3-day meal plan”
- “Fix this resume for a marketing job”
Clear instructions, context, and feedback make AI work better, just like with a person.
Why prompts matter for work
Used well, AI can take on a variety of roles:
- An extra brain to bounce ideas off of
- A junior employee to handle grunt work
- A consultant for tasks you’ve never done before
See more examples under “What You Can Actually Use AI For”.
You’ll work faster, improve quality, and free up time.
What tools are out there?
Here are some popular ones (all use prompts, so skills transfer):
- ChatGPT: Well-known, all-purpose. GPT-4 is best.
- Claude: Similar to ChatGPT, strong for writing.
- Grok: Built into X, fast, good for real-time info.
- Perplexity: Great for research, combines search + AI.
- DeepSeek/Gemini/Mistral: Newer models with unique strengths.
Each has different pricing, strengths, and weaknesses. Experiment to find what suits you. I often switch based on the task or performance. I find some models do better at certain tasks, and some feel slower on certain days, so I’m not that rigid with what I use when. Sometimes I ask two models the same thing and cherry pick from their responses.
Common prompting mistakes & fixes
Prompting improves with practice. You don't need perfect or premade prompts. Premade prompts are useful references but you don't want to copy and paste every time. Instead you'll want to type as you think, reacting to outputs in realtime.
Here's how to make your prompts effective:
1. Vague prompts = vague results
“Write a reply to this email” is like saying “make food.” Ok… what kind? for who? when? how much?
✅ Fix: Give it direction. Tell it what you want, who it’s for, what tone to use, and what not to do. Be specific. Ask for bullet points, a casual tone, or a 2-line tweet.
The more specific, the better the results.
2. Not adding enough context
ChatGPT isn’t a mind reader. If you say “fix my resume” but don’t give it your job history or what the role you want, you’re wasting your time.
✅ Fix: Give it the full picture. I like to mention that I’m going to provide examples or context and then just paste everything in at the end of the prompt.
The good thing with AI is that it can handle a lot more information than a regular person. With a human you might try to be thoughtful about how much information you dump on them, or format it in a digestible way. With AI you can send it all.
The only time it’s too much is if it’s not relevant to what you asked for and confuses the AI.
3. Asking for too much at once
“Make me a blog post, 10 headlines, a social caption, and an email…” Asking for 10 things at once is a guaranteed mess. You’ll get weak or incomplete results.
✅ Fix: Split it up.
- Start with: “Give me 5 blog title ideas.”
- Then: “Write an outline for title #3.”
- Then: “Now write the intro.”
If you’re not sure how to break it down you can just ask it to do that as well. AI models are generally eager to take action so you might need to tell it to hold up until all the details are defined.
An example is: “I want to market our new feature. I need to write a blog post, a few social posts, and reach out to content partners. Split up the tasks so that we can tackle them individually. Detail out what we need for each step. Just give me a plan, don't start yet.”
4. Not specifying tone or style
Wanted to write a social media post but it sounds like a b2b saas pitch?
✅ Fix: Tell it what tone you want.
- “Make it sound like a Gen Z creator”
- “Use a casual tone with lowercase that doesn’t sound like it’s trying too hard”
If you want specific results, it helps to paste in examples. For example when getting help drafting tweets, I'll paste a bunch of my previous ones as reference.
I mention it below but tone and writing style is still a mixed bag with AI so even with great prompting you’ll probably end up editing it a bit yourself.
5. Not checking the output
Even good models make stuff up or miss the tone. It might mostly be good and have some parts that are totally off.
✅ Fix: Treat it like a draft.
- Skim for errors
- Reword to match your voice
- Add missing nuance
Use AI to get 80% there. Then make it yours.
💡 One way to instantly improve your prompts
This has been extremely effective for me.
Instead of stressing about getting everything right in the first go, I just ask the AI to ask me questions.
At the end of your prompt just add “Ask me clarifying questions before your start.”
Why it works:
- It’ll check assumptions before responding
- It’ll surface anything that might be missing
- It turns the interaction into a dialogue, not a one shot
This helps even if your original prompt is messy or incomplete.
Key Takeaways
- Great prompts need clear outcomes, context, and constraints
- Use prompts like building blocks. Stack steps together.
- Use AI like a coworker. Let it clarify and push back.
- Don't overthink it. The more you use it, the better you get.
🚀 What you can actually use AI for
So what can you actually do with prompts?
A few ways people use it every day:
Thinking partner
- Brainstorm ideas for a blog post
- Explore different angles for a marketing campaign
- Work through a half formed business idea via conversation
- Pressure test assumptions before investing time
It helps you think with someone, even if you're working alone.
Productivity tool
AI is great for offloading manual labor:
- Summarize dense articles or meeting transcripts
- Clean up and reformat data or notes
- Generate outlines, bullet points, or draft versions of content
- Write routine updates, reports, or emails based on your input
You focus on the high level stuff. Let the AI handle the grunt work.
Expert consultant
It’s useful for finding information catered to your situation.
- Learn how to file quarterly taxes as a freelancer, including deductions like home office expenses, without needing a CPA.
- Optimize a landing page by targeting specific keywords and using A/B testing based on best practices
- Draft a Terms and Conditions agreement for your mobile app, covering user licenses and prohibited activities, without paying $500/hr for a lawyer.
Is it perfect? No.
But it’ll get you 80% of the way. It’s faster and cheaper. You can always fill in the gaps with a real expert after.
Second opinion
Working alone? It’s hard to see your own blind spots.
AI can:
- Critique your writing
- Challenge your thinking
- Spot weak arguments
- Suggest missing angles or counterpoints
- Call out your blind spots based on your conversation history
It won’t replace thoughtful peer feedback, but it gets you a lot further than staring at your own doc for hours.
Personal assistant
AI doesn’t just just have to be for work. It’s useful in personal life too.
- Plan your meals for the week
- Create a custom gym routine
- Draft a message to your landlord
- Prep a checklist for filing your taxes
- Talk through your feelings
Anything that’s a bit annoying or repetitive?
There’s probably a way to offload part of it to AI.
🧠 Building AI workflows
The trick to doing all the above is to use AI through the whole process instead of trying to get it do achieve it in one prompt. Have a conversation.
This section is about that process:
How to break down a goal, structure the steps, and collaborate with AI to get useful results.
Step 1: Check task size
Some tasks can be accomplished in one prompt, others need a more interactive approach. I often get too ambitious and try to one shot a task and I get better results if I break it up.
- Small task? e.g., "Rewrite this email." → One prompt.
- Big task? e.g., "Plan a product launch." → Split up tasks, use multiple prompts via conversation.
Step 2: Break it down
For big tasks, set context and break it down. You can ask AI or do it yourself if it’s a familiar task.
- "I want to build an app. Break it into stages: idea → validation → design → build → ship. Give me an overview of what I need to do for each stage."
AI gives a roadmap. Then tackle each phase. Works for blogs, projects, hiring, etc.
Step 3: Focus on each step
Work step by step with specific prompts:
- "Draft a blog post from this outline."
- "List interview questions for this role."
Stay narrow for better results, aligning with the big goal. This gives you a chance to review progress during.
Step 4: Review & refine
AI delivers drafts, but you’re the boss.
- Check for errors or weak spots.
- Refine with feedback, e.g., "Make the intro more casual."
Own the final output through iteration.
Essentially you’re delegating to AI across domains whether you’re writing, building, planning, or replying.
- Set the goal
- Break down the task
- Use AI to help with each piece
- Give feedback
- Own the final result
If you’re a manager or run a start up you’ll probably find this pretty intuitive. If not don’t worry. The more you practice this process, the more second nature it becomes.
AI works at different levels of thinking
You can use AI at different layers of thinking required at different stages of a task. Knowing this can help you think of new ways to use it.
-
High level
Planning, scoping, ideation, and strategy. -
Mid level
Break down tasks, identify components, and create outlines. -
Low level
Execution: writing, formatting, generating, and editing.
tldr; Better communication = better results
Organize your work into digestible pieces and work through it step by step.
The reason this helps is because often the main issue is the communication gap between you and AI. Just like working with people, transferring all the context and intent that’s in your head can be difficult. Even if AI was perfect if you aren’t on the same page then you won’t get good results.
Splitting up tasks makes it easier to ensure that you’re properly communicating everything you want to the AI.
🔄 Real life workflow examples
Here are 4 real examples of where you’d use AI to get something done in multiple prompts.
✍️ Workflow 1: Creating Content
Goal: Plan and write content that grows your audience.
Step 1: Set the content goal + context
Prompt 1:
“I’m planning content for LinkedIn to grow my audience.
My niche is helping beginners get better at AI.
I want this post to make people realize they’re using ChatGPT wrong, and teach them a trick to instantly improve their prompts.
Don’t write the post yet — just help me clarify the hook and structure. Ask clarifying questions”
AI helps shape the direction, suggests angles (e.g. “quick win” vs “frustration with results”), and offers possible structures like:
- Story > Insight > Action
- Expectation > Reality > Takeaway
It’ll also ask questions to solidify its understanding of the task.
Step 2: Draft and refine
Prompt 2:
“Let’s go with the quick win angle. Use a confident but not gimmicky tone — like a sharp Twitter thread. Maybe we can mention “example idea” or “example idea”
Here are a few examples of posts I’ve written. Match the tone.”
AI gives you 2–3 versions.
Then you say:
“Option 2 is closest. I like the rhythm of the first two lines, but the ending is too flat.
Can you keep the structure but land it with more punch?”
You go back and forth until it clicks.
Step 3: Support with visuals or a content plan
Now that you have the post:
- Ask AI to brainstorm 3 ideas for supporting visuals
- Generate 5 headline/title variations
- Ask: “What other posts could I write that build on this one?”
Workflow summary:
Instead of drafting a generic post you:
- Aligned on the goal and message
- Iterated to land on the right tone and flow
- Generated assets to support or extend it
🛠 Workflow 2: Turning a Product Idea Into a Spec
Goal: Turn a half baked feature idea into a shareable, structured product spec.
Let’s say you’re a PM with a new idea in your head. Maybe it ties into an engagement goal or came from user feedback. You want to move forward with it, so you need to flesh out the details.
Step 1: Brainstorm and refine ideas
Prompt 1:
“I’m a PM working on a social app that does X. I want to build a new way for users to react to photos.
I’ve thought about emoji reactions, freeform doodles, or even audio replies.
The goal is to boost engagement and give creators feedback.
Help me explore these ideas, and challenge my thinking.”
You can dump as many thoughts and ideas that you have at this point.
AI asks helpful questions (e.g. moderation, latency, UX friction), and surfaces considerations you might’ve missed.
Step 2: Generate a first draft spec
Prompt 2:
“Let’s go with the audio replies idea.
Can you create a first draft of the product spec, including:
- User flow
- Key edge cases
- Data we’d need to capture
- Anything that might break?
You get a solid first version to react to.
Step 3: Zoom in on details
Now that the spec exists, you can:
- Ask: “What if someone hits record but closes the app?”
- Ask: “List all possible states for this UI component”
- Ask: “What metrics would we track to evaluate this feature?”
AI helps you think through UX, edge cases, and instrumentation.
Step 4: Prep for sharing and implementation
Once you’re happy with the draft:
- Ask for a short summary for stakeholders
- Ask it to write an implementation checklist for engineering
- Ask for key open questions to raise in your next product meeting
Workflow summary:
Here we used AI to:
- Explore and refined ideas
- Create a draft spec
- Work through product edge cases
- Get outputs to share or build from
🏃 Workflow 3: Planning for a Personal Goal (Run a Marathon)
Goal: Go from “I want to run a marathon in October” to a structured, realistic training plan.
Let’s say you have a big goal in mind but you don’t know where to start. What do you focus on or avoid burning out?
Step 1: Set the goal and let AI ask questions
Prompt 1:
“I want to run a marathon this year.
I’m not sure where to start.
Act like a coach. Ask me the right questions, then help build a plan.”
AI asks stuff like:
- When’s the race?
- What’s your current fitness level?
- How much time can you train per week?
- What’s your goal: just finish or hit a specific time?
You respond, and it maps out the big picture.
Step 2: Build the phases + milestones
Based on your info, AI gives you:
- Training phases (base, build, taper)
- Weekly structure (long runs, cross training, rest)
- Monthly milestones (distance/time targets)
- Checkpoints to reevaluate progress
Step 3: Expand into nutrition, gear, recovery
Prompt 2:
“Create a basic nutrition checklist for long run days.
Also give me reminders for recovery. What should I do weekly to prevent injury?”
You get:
- Fuel and hydration tips
- Recovery protocols
- Pre-race prep reminders
Step 4: Stay accountable + adapt
Prompt 3:
“I want to check in with you every Monday.
I’ll tell you how last week went, and you update the plan based on progress.”
You’ve now got a lightweight, custom coach that evolves with you.
Workflow summary:
Here we used AI to:
- Define your goal
- Build a realistic, phased plan
- Cover the full system (training, food, recovery)
- Set up ongoing accountability
🗣️ Workflow 4: Turn Raw Feedback into Clear Action
Goal: Process a wall of customer feedback, pull out insights, and figure out what to do next.
Let’s say you’ve just dropped 40+ pieces of raw feedback into a doc from surveys, help tickets, and comments in your Discord. You want to know:
- What’s standing out?
- What’s broken?
- What should we do about it?
AI is great for this.
Step 1: Summarize the mess
Prompt 1:
“I’m going to paste a bunch of unstructured user feedback from different sources.
Can you help me group them into themes and summarize what people are saying for each?”
You paste the full dump.
AI returns a summary like:
- Onboarding confusion: multiple users didn’t understand how to start their first project
- Pricing concerns: a few people asked if there’s a free tier or trial
- Feature requests: custom reminders, better image uploads, team sharing
You’ve saved an hour of mental effort.
Step 2: Find action items
Prompt 2:
“Based on that summary, what are the clearest next steps we could take?
Flag anything that’s directly on me or where we’re blocked.”
Of course mention anything you think is relevant here like your goals.
AI returns:
- Add onboarding tooltips or a quick start checklist
- Clarify pricing on the homepage
- Follow up on the upload bug. You were tagged in that ticket
- Discuss feasibility of team sharing in next product meeting
Now you can spend more time on solving problems instead of processing.
Step 3: Delegate back to AI
You can think through these and decide which to pursue. Now for each item, you can ask:
- “Can you write copy for the pricing section that better sets expectations?”
- “Draft a message to the Discord group explaining how reminders work now + asking for feedback on proposed changes”
- “Help me prep questions to bring to the product team about team sharing. What should we validate before building it?”
AI helps you figure out what to do, and then helps you do it.
Step 4: Make a decision
Let’s say one piece of feedback sparked a strategic question:
“Some users want more control, but others want it simpler.
Should we add more settings or simplify the defaults?”
Prompt:
“Help me weigh the tradeoffs between adding more advanced settings vs keeping things minimal.
Give me 3 pros and cons for each path, and suggest when one makes more sense than the other.”
Workflow summary:
You used AI to:
- Understand what users are saying
- Identify what needs attention
- Delegate specific tasks
- Make decisions based on real insight
🧭 Before You Start: A few more tips
The next step is to try some workflows. Have a chat with AI to see how it can help you.
Here are just a few heads ups before you start. Things that tend to confuse or frustrate beginners and even experienced users sometimes.
1. Sometimes AI Gets Stuck. That’s Normal
Sometimes no matter how many ways you try to explain something, the AI doesn’t get it.
It might:
- Repeat itself
- Miss your point
- Keep giving safe, generic answers
✅ What to do:
- Start a new chat. This resets the context and often fixes weird behavior.
- Reword your prompt. Even small phrasing changes can help.
- Try another model. Claude, Grok, Gemini, and others all handle tasks slightly differently. Sometimes switching gives you a better result.
Pro tip: If something isn’t working after a few tries, don’t brute force it. Reset and try again.
2. You Can (and Should) Give Feedback Mid Conversation
If you’re not getting what you want, tell it.
“This isn’t specific enough.”
“Try again, but make it more visual.”
“I like the first half of this, but not the ending.”
“Tone is too formal. Rewrite it more casual.”
AI responds well to direct, specific feedback.
You don’t need to be polite or technical. Just say what you want.
3. If you're using AI to write, watch for these giveaways
AI still sounds like AI. If you're getting text for public facing content, watch out for these.
Common AI words and phrases
There are words and phrases that AI uses more frequently than humans. Usage of words like “delve” and ”tapestry” have exploded since ChatGPT launched. It also likes cliched phrases. Keeping these in your writing is a a dead giveaway.
“Delve into the possibilities…”
“Navigate the ever-evolving landscape…”
✅ Fix: Check the output against lists of overused words and phrases. Rewrite in your voice.
Over hyphenation & long dashes
AI follows exact spelling and that means it hyphenates compound words: “high-traffic”, “fast-paced”, “short-form”. It might be technically correct but people usually don’t hyphenate these. Iit also likes to break sentences with dashes - even when it doesn't seem necessary.
✅ Fix: Replace dashes with spaces and split into shorter sentences. A good rule of thumb: would you hyphenate it?
Lists of three everywhere
AI loves to use list of three: “readers, clients, or friends” or “lazy, sloppy, and cheap”. Nothing wrong with it inherently, but once you notice the frequency it looks unnatural.
✅ Fix: Keep the strongest example. Ditch the others. Change up the number of examples.
Wordy
AI rambles. By default it says in 20 words what could’ve taken 7. This is especially noticeable in tweets, emails, or anything short form.
✅ Fix: Ask for “punchy” or “concise.” Trim without mercy.
This isn’t an exhaustive list but covers the most common issues. This type of list will continue to change as AI evolves. The more you use AI, the more you'll notice these patterns intuitively. Keep this list in mind to draft rapidly while still producing work that feels authentic to you.
🎉 You're Ready
You’ve learned how to prompt better, what AI can help with, and how to build effective workflows.
Don’t overthink it and just try it!
You’ll get a better feel for it over time and you’ll keep finding new ways to use it. These models are also improving quickly and getting better at understanding what we want.
AI won’t be perfect. Don’t expect it to deliver finished work on the first go. But it will help you move faster: drafting, iterating, and processing ideas much faster than on your own. Let it handle the heavy lifting while you guide the work.
Keep It Flexible & Get Creative
I need to mention you don’t need a rigid 5 step process to use AI well. There’s no rules. What I shared are just guidelines to make it easier. The real key is being able to clearly articulate what you want, share enough context, and give feedback to direct the AI. Do that and you’ll be fine.
Also, get creative. Don’t limit yourself to tasks you’d ask a colleague or employee. Try anything. Ask wild questions, experiment, see what happens. You might have blind spots where AI can help, especially in areas you’d normally tackle alone. Just ask stuff. The worst case is you learn what doesn’t work. This technology is so new, half the fun is figuring out what you can use it for.
🔁 What’s Next?
Need inspiration?
Try using AI to get something done — pick a task and break it down
Explore our prompt library — packed with ready to use prompts by use case
Check out the blog — we post AI tips and learnings regularly
Follow on X to stay sharp as the tools evolve