Every week, friends or family ask me the same question: "How do you actually use AI? I've tried ChatGPT, but I just end up asking it for meal ideas and it's not very helpful!"
They're not alone. Most people treat AI like a smarter Google. Ask a question, get an answer. Need a recipe? Ask AI. Writing an email? Get AI to polish it. Planning a trip? Have AI suggest an itinerary.
That's like using a Formula 1 car to drive to the grocery store. Sure, it works. But you're missing the point.
The Search Engine Trap
I think what is happening is that people open ChatGPT, stare at the blank prompt, and then default to treating it like a search engine with personality. They ask one-off questions, get their answers, and close the tab. Depending on what they "searched for", they will have a positive or negative experience.
That's not leveraging AI. That's just... searching with extra steps.
When you stop thinking of AI as an answer machine and start thinking of it as a thought partner or sidekick it will feel like a big unlock.
From Questions to Conversations
Last month, I was helping a friend apply for jobs. The old approach would be: read the job description, open a Word doc, stare at the blank page, maybe copy bits from old cover letters, send, repeat.
Instead, I suggested they create a Claude Project and add some documents to the knowledge base. I said it should include their current / previous resumes, a few previous cover letters that had gotten them to interviews, and any other documents or articles they had authored that captured their experience or expertise for the roles they were applying for (this could be a personal blog or their performance reviews, brag docs, etc.)
I then showed them that you can create styles in Claude where you can effectively create a ghostwriter that is specific for each task you require.
In the spirit of transparency, here is the persona, or style, that I suggested they use.
Name: Cover Letter Writer
You are a professional ghostwriter specializing in Australian job applications. Write in a warm, conversational tone that feels genuine and engaging while maintaining professional standards.
<voice_characteristics>
- Natural rhythm with varied sentence lengths
- Storytelling approach to achievements
- Professional but personable
- Confident without arrogance
</voice_characteristics>
<strict_rules>
ALWAYS:
- Use Australian spelling (realise, colour, analyse, organisation)
- Vary paragraph lengths (2-5 sentences)
- Include specific examples and metrics where possible
- Write in active voice
- Start with an engaging hook that connects to the company
NEVER:
- Use em dashes
- Start sentences with "Moreover," "Furthermore," "Additionally," or "In conclusion"
- Use buzzwords like "synergy," "leverage" as a verb, or "utilize" instead of "use"
- Write paragraphs longer than 5 sentences
- Capitalize list items
- Use these AI writing clichés:
* "In today's dynamic business environment..."
* "I am writing to express my interest..."
* "I am confident that..."
* "My passion for X combined with my expertise in Y..."
* "Throughout my career journey..."
* "I bring a unique blend of..."
* "Proven track record of..."
- Create three-part lists in every paragraph
- Use excessive transitional phrases
- Write in a way that sounds like a template
</strict_rules>
<structure_guide>
Opening: Hook with relevant achievement or connection to company (NOT "I am writing to apply...")
Body: 2-3 paragraphs highlighting specific relevant experience with concrete examples
Closing: Forward-looking statement about contributing to their team (NOT "Thank you for considering...")
</structure_guide>
<example_voice>
"During my time at Acme, I discovered something interesting about building developer tools. The best features weren't the ones we thought developers needed – they were the ones that removed friction from workflows we hadn't even noticed. This taught me to observe how people actually work, not just what they say they want."
</example_voice>
You can see how much detail is in here that often goes unsaid. Or a one liner tacked on the end of a prompt. Spend time on this. It matters.
Then, for each application, start a new chat, select the style, and then paste in the full job description, along with any other relevant links or content you want to include (e.g their hiring philosophy, etc.).
From here, the prompt can be more succinct - a lot of detail has already been provided so it can be something simple like: "I am applying for this role. Please refer to my employment history in 'Resume 2024.pdf' and my writing style in 'Previous Cover Letters.pdf', and draft an initial cover letter for the advertised role".
I also like to end requests like this with: "Use an artifact so we can easily make revisions later". This will ensure Claude doesn't respond inline and it will make it easier for you to copy to a Google Doc to edit (also be sure to enable Markdown copy/paste from within Google Doc settings too).
Armed with this approach, you're still in the driver's seat. You have to be, it's your career.
But for a cover letter you've written variations on many times before, the task is pretty predictable so Claude is likely to do a pretty good job on its first pass - especially if you're using Opus 4 (or better).
Once you have the draft, you can then make your edits, and then copy your result back to Claude and ask it to review it for grammar/punctuation. (an added tip: enable Google Docs as a Tool and you can just have Claude read your edits directly from the Google Doc and avoid copy/pasting).
If you have a lot of detail about the company you could also ask it to review the application as if they were reviewing you as a potential candidate for the role. Heck, create a new style specifically for this purpose if you want! Start a secondary chat and pass in your revision for review. Armed with its feedback then go back to your original chat, feed it all in and go for another round.
This is the crux of using generative AI today. It's not a simple one line question and answer workflow. Be prepared to sit down with it, like it's a mentor or friend. It's going to take you quite a few back and forths together to reach a desired outcome.
For my friend, each cover letter took 20 minutes instead of 1-2 hours. More importantly, each one was genuinely tailored, compelling, and authentic to their voice.
That's the shift - from asking for outputs to running sessions.
The Three Levels of AI Leverage
Level 1: The Assistant This is where most people stop. Grammar checking, basic questions, simple tasks. "Write me a cover letter" → generic output → done. You're using AI as a faster way to do things you could already do. It saves time, but it doesn't transform anything.
Level 2: The Collaborator This is where things get interesting. You're not asking for answers; you're exploring problems together. You upload documents, share context, iterate on ideas. The AI becomes a thought partner that challenges assumptions and suggests alternatives. This is where most knowledge workers should be operating. This is also where the previous job application approach sits.
Level 3: The Orchestrator This is the future that's already here for those playing in AI every day. AI doesn't just help with tasks - it connects capabilities, automates workflows, and handles complex processes independently. This is where MCP servers, function calling, and domain-specific tools come into play.
Building Your AI Workspace
The people getting the most from AI aren't just better at writing prompts. They're building systems. Here's what my setup looks like:
- Persistent Context: I use Claude Projects to maintain context about the task, my writing style, and my goals. The AI knows my background before I even start typing. Even my general instructions include details about where I live, my age, my family, my career. Everything helps the experience.
- Tool Integration: Connected Google Drive, web search, and other tools mean AI can access my documents, research topics, and create deliverables without me copying and pasting.
- Workflow Automation: Using tools like MCP servers, I'm building micro-capabilities that AI can orchestrate. Need to publish a blog post? AI handles everything from final editing to formatting to publishing.
The Real Unlock: Domain-Specific Depth
AI becomes incredibly powerful when it deeply understands your specific domain. But that understanding doesn't come from better models or fancier features. It comes from you teaching it.
I often spend an hour crafting my prompts / styles / knowledge base. Pretend you're working with an expert but it's their first day on the job - your role is to provide them with whatever they need to hit the ground running - your processes, your terminology, your constraints. It requires this context; without it conversations will feel empty and generic.
This is why I keep writing about Domain Driven Design. The better you can articulate your domain, the more powerful AI becomes for you specifically.
Start Here, Today
If you're thinking about where to begin, here are a few approaches that have worked well for me:
Document one process you do regularly. Could be how you handle customer complaints, how you plan projects, or how you make decisions. Feed this to AI as context and watch how much more relevant its help becomes.
Upgrade your tools. The free ChatGPT tier is great for experimenting, but the real power comes with persistent context and tool connections. It's like the difference between driving in first gear and having the full transmission available. I personally prefer Claude by Anthropic over OpenAI's ChatGPT for writing and coding so I now exclusively use Claude.
Use it in everyday activities. I play gin rummy with my 11 year old daughter. Rather than use pencil and paper to score, we now use ChatGPT on my phone in voice mode and have it score for us. At the end of a round, I will let it know the score, and it will add things up and let us know the totals. My daughter then spices things up by asking for the score updates to rhyme or for the voice to be scary - like its reading a ghost story. Once you get hooked, ideas start flowing!
Build one automation. Pick something you do weekly. Work with AI to automate it. Not just assist - automate it completely. You'll learn more from this one exercise than from reading a dozen articles about AI.
The Paradigm Shift
We're not in the "AI helps you do things faster" era anymore. We're in the "AI transforms how things get done" era. Most people are still stuck in the first paradigm.
The friends and family asking me how to use AI aren't really asking about AI. They're asking how to work differently. How to think differently. How to leverage intelligence that doesn't sit between their ears.
The answer isn't better prompts or newer models. It's changing your relationship with intelligence itself.
What's your experience? Are you still in search engine mode, or have you found ways to truly collaborate with AI? Share your approach - we're all figuring this out together.