Most people use AI the same way they use a search engine - ask a question, get an answer, move on. But the people getting the most out of it are doing something fundamentally different.
The Way Most of Us Are Using AI Right Now
There's a pattern that shows up constantly with new AI users. You sit down with a blank page, feel stuck, and think: "I'll just ask the AI to write this for me." You type something like "write me a product brief for X" and get back a clean, confident-sounding document that somehow says nothing useful at all.
The output looks right. The sentences are grammatical. The structure makes sense. But when you read it back, it doesn't reflect your actual thinking - because you never put your actual thinking in. The AI filled the space with plausible-sounding words, not with your specific context, your real constraints, your genuine instincts about the problem.
This is the core frustration that many product managers, freelancers, and small business owners hit within the first few weeks of using AI tools. The outputs feel generic. They don't capture the nuance of your situation. And so people either give up on AI entirely or keep using it for shallow tasks - proofreading, reformatting, summarizing - while still doing all the hard thinking alone.
The Shift: From Dictation Tool to Thinking Partner
The change happens when you stop treating AI as a place to dump tasks and start treating it as a collaborator in your thinking process. This is less about prompting technique and more about mindset.
When you ask AI to produce something, you're outsourcing the cognitive work. When you ask AI to think with you, you're using it to sharpen your own thinking. The difference in how you talk to it is subtle but significant. Instead of "write a brief about X," you say something like: "Here's what I'm trying to accomplish and here's what's not working - help me figure out what I'm missing." You're bringing your incomplete, messy thoughts into the conversation and asking the AI to push back, ask questions, or offer a different frame.
This works because good AI systems are trained on an enormous range of problems, disciplines, and ways of structuring ideas. When you share the context of your actual situation - the real constraints, the stakeholder tensions, the things you've already tried - the model can draw on patterns from adjacent fields or surface assumptions you haven't questioned. It's not magic. But it is a genuinely useful second perspective, available instantly, at any hour, without the social overhead of pulling a colleague into yet another meeting.
Real Example - Step by Step
Say you're a product manager trying to write a roadmap justification for a feature your engineering team is skeptical about. You've got a rough idea but you can't get the argument to land.
Step 1: Open with your real problem, not a task. Don't say "write a roadmap justification." Instead: "I'm trying to justify prioritizing a customer notification feature. Engineering thinks it's low complexity but not impactful. I think it reduces churn but I can't articulate why clearly. Help me think through this."
Step 2: Let the AI ask clarifying questions. A good AI response here won't just hand you a document. It might ask: What does the churn data actually show? Is there qualitative feedback from support tickets? Have customers left because of this specific problem or something adjacent? These questions are the value - they expose the gaps in your own reasoning.
Step 3: Answer the questions out loud (or in text). Your answers don't need to be polished. Type out your reasoning, even if it's half-formed. "We have three support tickets this month about missed updates. I think there are more but we're not tracking it well." That raw material is what the AI uses to help you build a sharper argument.
Step 4: Ask for the counterargument. Once you have a draft logic, explicitly prompt: "Now argue against this. What would a skeptical engineer or CFO say?" Getting the steelman version of the opposition helps you stress-test before the real conversation.
Step 5: Now ask for the document. At this point, writing it up becomes almost mechanical - because the thinking is done.
How to Apply This Today
The next time you feel stuck on something - a proposal, a strategy doc, a difficult email - try this before you write a single word: open an AI chat and describe the problem out loud. Not the deliverable. The problem.
Tell the AI what you're trying to achieve, what's blocking you, and what you've already considered. Then ask it to help you think, not to produce output. You can always ask for the polished version later, once the thinking is solid.
If you manage a team, consider introducing this approach explicitly. When someone says they're stuck, suggest they spend 15 minutes in an AI conversation before they ask for a meeting. Not to replace collaboration - but to arrive at that collaboration with sharper questions already formed.
And if you're skeptical that AI can add value to your specific domain - test the collaborative mode before you judge it. The generic-output problem that most people experience is real, but it's mostly a function of how the question is asked.
Key Takeaways
- AI gives generic output when you give it generic input - context is everything
- Treating AI as a thinking partner means bringing your messy, incomplete reasoning into the conversation, not polishing it first
- Asking for counterarguments or probing questions is often more valuable than asking for a finished document
- The goal is to use AI to sharpen your thinking, then write from a place of clarity
- The best outputs come at the end of a conversation, not the beginning
What's your experience with this? Drop a comment below - I read every one.
Sources referenced: HackerNews discussion thread on Claude, Anthropic model release coverage
Top comments (0)