Try this experiment. Open any AI chat. Type: "I'm thinking about a career change. What should I do?"
You will get something like: "Consider your strengths and passions. Update your resume. Network in your target industry. Explore upskilling opportunities. Maybe try freelancing on the side."
Generic. Safe. Useless. You already knew all of that.
Now try a different approach. Type: "I'm a mid-level developer at a 50-person company earning $95k. Here are four paths I have already considered: (1) stay and negotiate a team transfer, (2) take a startup offer at $80k + equity, (3) stay and build side projects for 6 months, (4) job hunt for mid-size companies. My criteria are learning velocity, financial floor of $70k, and commute under 40 minutes. What am I not seeing?"
Completely different output. Specific. Tailored. It evaluates your actual options against your actual constraints and surfaces things you missed.
Same tool. Different input. Radically different quality.
The Problem: AI Pulls from Averages
When you ask AI to generate options from scratch, it does what any language model does: it predicts the most statistically probable response. That means average advice. The kind of thing most people would say. The median response.
Average advice is not bad advice. It is just not your advice. It does not account for your specific constraints, your risk tolerance, your market, your timeline, or the things that make your situation unique.
AI trained on the entire internet has seen millions of career transitions. But when it generates options without your input, it defaults to the most common patterns. The ones that work for the average person. Not the ones that work for you.
The Fix: You Generate, AI Evaluates
The solution is a division of labor. You do what you are good at: knowing your situation. AI does what it is good at: analyzing tradeoffs across multiple variables.
Here is the pattern:
- You list 3 to 5 options you have already considered. Even the obvious ones. Especially the ones you have been going back and forth on.
- You state your criteria. What actually matters to you, not what should matter.
- AI evaluates each option against your criteria. It finds tradeoffs you missed. It flags risks you underestimated.
- AI identifies blind spots. Things you assumed were true that might not be.
Notice: AI never invents the options. You do. AI's job is to evaluate, not originate.
Why This Works
When you give AI your options, you give it constraints. Constraints are the secret to good AI output. Without constraints, AI has infinite degrees of freedom and defaults to the safest, most average response. With constraints, it has something specific to push against, and the output becomes sharp and relevant.
Your options are constraints. They tell AI: "Don't give me the universe of possibilities. Work within these five paths and tell me which one fits my situation best."
Think of it like hiring a consultant. You wouldn't walk into a consulting meeting and say "what should I do with my business?" You would say "here are three strategies I am considering. Here is my context. Help me pick the right one." The same principle applies to AI.
The Common Mistake
People skip this step because generating options feels like work. It is easier to type "what should I do?" and let AI do the thinking. But that abdication is exactly what produces bad output.
The five minutes you spend listing your options pays for itself ten times over in the quality of AI's analysis. Because AI is not replacing your thinking. It is amplifying it. And it can only amplify what you give it.
What This Means for Strategic Prompts
If you take one thing from this article, take this: never start a strategic AI prompt with "what should I do?" Start with "here is what I am considering. Help me choose."
The difference is not subtle. It is the difference between a tool that wastes your time and a tool that changes how you make decisions.