How I used AI as a sparring partner for high-bar interviews in tech?
My 4 step approach on using AI for mock interviews in tech. AI can structure your answers, but the magic happens when you think and speak for yourself
⚠️ Disclaimer: If you think you can just read an AI-generated script and ace an interview, this post isn’t for you. Real prep takes effort, iteration, and reflection.
Long before my career in tech, I spent 7 years in Model UN and debate—where I learned the power of saying ideas out loud and thinking on my feet.
Back then, I practiced in front of a mirror. Today, I practice in front of ChatGPT (or another model of choice. Claude is pretty good with words)
This article is about using AI as a sparring partner to help you sound clear, structured, and genuinely you.
My 4-step system
Step 1: Create a “Sh**ty First Draft” for ChatGPT to polish
🧠 Write your messy answer to a questions in 10 minutes. Just get it out by brain dumping then process it through AI for polishing.
Here is a prompt template:
“Here’s my answer to [interview question] for [job description] at [company]. Critique it with a high bar and rewrite using STAR method.
Think step by step and make sure to
1. Apply the pyramid principle (top line summary first to set the stage)
2. Check for MECE in my actions
3. Suggest where my story is weak”
Step 2: Review and refine the answer from by first analyzing the output STAR answer:
Did it clarified your situation?
Which actions did they emphasize? Are those correct?
Were the quantified results accurate and impactful?
Now leverage role-based prompting and provide additional context if the polished version seems off. This forces you to see your answer through the eyes of a high-bar evaluator, identifying gaps and blind spots you might miss on your own.
Here is another prompt to try:
“Here [XYZ] is what I think went wrong. My actual action/result was [ABC].
Knowing that, what would a senior leader in. [target function] at [target company] evaluate my response if they were interviewing me?
How should I improve the answer to increase chance of this leader hiring me for the [job description]? ”
Step 3: The verbal test 🎤 Now the real work: speak your answer out loud to ChatGPT’s voice feature.
Say your answer out loud and record it (or use speech-to-text). First, listen critically to yourself:
Where does the logic falter?
Are actions or results missing or redundant?
Where could it be tighter or more compelling?
Does this truly sound like you?
Take notes and think through improvements before looking at AI feedback.
Then, ask AI to analyze the recording or transcript — the AI may highlight additional blind spots, but the core skill is training your own brain to detect and fix issues proactively.
Consider using different roles as target audience depending on different levels of interviewer might meet (cross-functional vs. direct leadership vs. skip level vs. peer)
Speaking is muscle memory. This is where breakthroughs happen. You’ll hear every “um,” every unclear transition, every spot where you lose the thread.
Step 4: Iterate until it’s yours 🔄 Loop steps 2-3 until the answer feels natural. Not memorized. Natural.
👉 By the third round, your answer should feel like something you’d say fluently in a real conversation—not like a monologue you memorized.
Once it’s ~80% ready, sleep on it. The next day, practice again cold, without looking at any notes. This pressure-tests whether one can communicate the story clearly on the spot without any assistance.
🧠 The Human Factor (non-negotiable)
AI can help structure answers, but it can’t think for you. That part has to be yours. Follow-ups like “What would you do differently?”reveal how deeply you’ve thought about your actions, trade-offs, and growth. My final tips:
Set the stage clearly. Give your interviewer a mental map before you dive in.
Don’t parrot AI. Translate GPT phrasing into your cadence and personality.
Use MECE to reflect. If they ask “What would you do differently?” you can show logical rigor by mapping gaps or overlaps in your original actions.
