Struggling to get a job? Acing the interview is critical.
Job interviews can be few and far between. When you get one, you need to walk in prepared.
As a coach, I’ve prepared my clients for job interviews for years, but recently I’ve found that an AI like ChatGPT or Claude can be the perfect practice partner.
Last week, I coached a client through doing exactly that. He had never thought of using AI for this purpose.
To illustrate without identifying him, suppose you’re a college senior who majored in linguistics and is interviewing at a tech company working with language models. This is such a specific and technical job that an AI might do a better job preparing you than any career coach.
Together, my client and I wrote a query describing the position and the questions he wanted to practice. We also asked for feedback after each round.
For our hypothetical linguistics graduate, it might look like this.
The query
I’m a college senior majoring in linguistics, interviewing for an entry-level role working with language models.
Here’s the job description, company information, and my résumé: [paste].
Act as the interviewer. Ask about my background, behavioral questions, and technical questions about applying linguistics to language models.
We’ll use audio. Ask one question at a time, wait for my answer, and ask follow-ups.
After five questions, stop and give specific feedback on where my answers were clear, vague, too long, or missing something important. Then let me try again.
What the conversation might sound like
AI: How would your linguistics background help you evaluate a language model?
Candidate: I’ve studied syntax, semantics, and pragmatics. That would help me recognize mistakes and understand why they’re happening.
AI: Consider “I saw the man with the telescope.” What are two possible meanings?
Candidate: Either I used a telescope to see the man, or the man had a telescope.
AI: How would you test whether a model can use context to distinguish them?
Candidate: I’d put the sentence into two stories. In one, I’m looking through a telescope across a harbor. In the other, I’m walking past people carrying scientific equipment. Then I’d ask who had the telescope.
AI: Would two correct answers establish that it can resolve this kind of ambiguity?
Candidate: Probably not. It might already know this example. I’d need to test other sentences too.
The feedback
AI: Your strongest answer described how you’d build a practical test. Your opening answer just listed subjects you’ve studied. Lead with what you can do with that knowledge.
You also needed prompting to acknowledge your test’s limitations. Include those yourself. Organize your answer around what you’re evaluating, how you’d test it, and what the results would let you conclude.
Once isn’t enough
The follow-up questions impressed me. The specific feedback impressed me more.
My client did four rounds, reading and listening to the feedback before trying again. By the fourth attempt, his answers were getting tight.
That’s why repeated practice matters. You need enough stories and potential answers in your head to speak with authority and fluidity. Knowing your subject doesn’t mean you can explain it when someone puts you on the spot.
I still had a role. I helped him set up the practice, ask for useful feedback, and put it to work.
AI vs. Humans
I’ve spent much of my career thinking about the relative roles of humans and machines in education and behavioral health. Coaches and therapists who insist that they will be better than AI in every situation have got it wrong. I just want my clients to succeed; I don’t really care if it is me or an AI, or most likely the combination of the two, that gets them there.
If my client leaves knowing how to practice without me, I’ve done my job. I don’t need to be the one asking every question.
