Interview Drills — Behavioral & Curiosity

Open-ended interview questions. Each has a Framework (the structure a strong answer follows), a Model answer (a concise example), and a Rubric (what an interviewer listens for). These are the behavioral, adaptability, and career-narrative questions from Topic 10 — interviewers use them to gauge whether you have genuine hands-on AI literacy, sound judgment about AI's limits, and the curiosity to keep up in a fast-moving field. Practice thinking aloud and always tie back to a concrete example. The app can role-play these as mock interviews (see mock-interview.md).

The universal behavioral structure (STAR): Situation (set the scene briefly) → Task (what you needed to do) → Action (what you specifically did, with detail) → Result (the outcome, quantified if possible) → and for AI-adjacent roles, a one-line Reflection on what you learned. Use STAR on almost any "tell me about a time…" question, and lead your career-narrative answers with the domain expertise + human skills + hands-on AI literacy combination.

D1

difficulty: easy
concept: landing-an-ai-role Walk me through an AI project you built — what it was, why, and what you learned.
Framework: STAR. Situation: the real problem you wanted to solve (ideally in your domain). Task: what "done" looked like. Action: what you actually built (the prompts/assistant/tool), the iterations, how you tested it. Result: the concrete outcome. Reflection: one thing the project taught you about working with AI.
Model answer: "As a paralegal, drafting client intake summaries ate an hour per case (Situation). I wanted a reliable assistant that turned raw intake notes into a structured summary (Task). I built a custom GPT with a tested prompt template, iterating about ten times and checking its output against twelve past cases to catch where it invented dates or missed conflicts (Action). It cut summary time to ten minutes, and I shared the prompt with two colleagues (Result). The big lesson: it sounded right far more often than it was right, so I built a verification checklist into the workflow (Reflection)."
Rubric: Strong answers describe a specific, real project with iteration and testing, and show the candidate evaluated the output rather than trusting it. The reflection reveals genuine hands-on judgment. Weak answers are vague ("I played with ChatGPT"), skip how they tested or improved it, or claim a flawless result with no learning.

D2

difficulty: medium
concept: evaluating-output Tell me about a time you caught an AI error before it caused a problem.
Framework: STAR. Situation: where AI output was about to be used. Task: your role in checking it. Action: how you caught it — what dimension you were checking (factuality, completeness, safety) and what made you suspicious. Result: what you did about it and the impact avoided. Reflection: how it changed your process.
Model answer: "I was reviewing AI-drafted help-center articles before publish (Situation). My job was the final quality check (Task). One article confidently cited a refund window of 60 days; it sounded authoritative, but it contradicted our actual 30-day policy, so I verified against the source doc (Action). I flagged it, we corrected it, and we avoided a wave of disputed refunds — then I found three more articles with the same invented number (Result). After that I started checking every policy figure against the source of truth and logged it as a recurring failure pattern for the team (Reflection)."
Rubric: Strong answers show the candidate looking past surface fluency to factuality, verifying against a source, and ideally spotting the error as part of a pattern — the core AI-Operations skill from Topic 5. Bonus for specific, documented feedback rather than just "it was wrong." Weak answers are generic, describe an obvious typo rather than a subtle confident error, or show no systematic check.

D3

difficulty: medium
concept: staying-current The AI field moves incredibly fast. How do you stay current?
Framework: State your system (the habits, not a list of headlines you chase) → emphasize hands-on experimentation over passive reading → give a concrete recent example of something new you tried and what you took from it → connect it to being "tool-aware, not tool-obsessed" (durable literacy over chasing every model name).
Model answer: "My main habit is hands-on: I keep a couple of small projects running so when a new capability ships, I actually try it rather than just read about it. I follow two or three reliable sources to know roughly what's newly possible, not every headline. Recently I tested a retrieval-style setup to ground answers in our own docs and saw hallucinations drop noticeably on internal questions — so I now reach for that pattern by default. I try to stay tool-aware without getting tool-obsessed; the durable skill is judgment about what AI can and can't do, not memorizing this month's model names."
Rubric: Strong answers describe a repeatable system centered on hands-on experimentation, name a specific recent thing they tried and learned, and show the maturity to value durable literacy over hype (Topic 9). Weak answers just list newsletters/influencers, claim to "read about AI a lot" with no hands-on element, or fixate on memorizing tool names.

D4

difficulty: medium
concept: landing-an-ai-role Why are you pivoting into AI from your current field?
Framework: Lead with the bridge, not a clean-slate story: name a genuine moment AI clicked for you in your existing work → frame your background as an asset (domain expertise + human skills) → show you understand the role realistically (working with AI, judging it) → end with forward-looking enthusiasm grounded in what you've already done.
Model answer: "I'm a nurse, and I started using AI to turn dense clinical guidelines into plain-language patient handouts — that's when it clicked that the hard part isn't generating text, it's judging whether it's accurate and safe, which is exactly what my clinical training is for. So I'm not starting over; I'm pairing eight years of domain judgment with hands-on AI literacy. I built a small portfolio evaluating AI medical summaries, and the more I do it the more energized I am — catching where confident AI is subtly wrong is work I'm genuinely good at and excited by."
Rubric: Strong answers frame the pivot as a bridge from real domain expertise, show the "I know this domain and I understand AI" combination (Topic 10), and convey authentic, specific enthusiasm backed by something they've actually built. Weak answers are generic ("AI is the future / it's exciting"), disown their past instead of leveraging it, or show no realistic grasp of what the role involves.

D5

difficulty: medium
concept: adaptability Tell me about a time you had to learn a new tool or skill quickly with little guidance.
Framework: STAR. Situation: the deadline or gap and why it mattered. Task: what you had to get competent at, fast. Action: how you learned — your experimental, learn-by-doing approach (the meta-skill for this field). Result: what you delivered. Reflection: how you'd apply the same mindset to AI's constant change.
Model answer: "Our team adopted a new analytics platform two weeks before a board report, with no training (Situation). I had to produce the report in it (Task). Instead of waiting, I rebuilt last quarter's report in the new tool, broke things, read docs only when stuck, and compared my output to the known-correct old numbers to validate (Action). I delivered on time and became the person others asked for help (Result). That experiment-validate-iterate loop is exactly how I approach new AI tools — try it on a known problem, check the output against ground truth, then trust it (Reflection)."
Rubric: Strong answers demonstrate adaptability as a learnable system — experiment, validate against ground truth, iterate — and explicitly connect it to thriving in a fast-moving AI field (Topics 6, 9). Weak answers describe passively waiting for training, show no self-directed approach, or never tie the example back to adaptability with AI.

D6

difficulty: hard
concept: ai-limitations-and-ethics Tell me about a time you disagreed with or pushed back on an AI-generated recommendation — how did you handle it?
Framework: STAR with a judgment lens. Situation: the recommendation and the stakes. Task: your responsibility to evaluate it. Action: why you doubted it (which risk — hallucination, bias, missing context, an unsafe edge case) and how you raised it constructively with evidence, not just instinct. Result: the decision and outcome. Reflection: the principle you now apply about when to trust vs. verify AI.
Model answer: "An AI hiring-screen tool ranked candidates and our team was about to filter by its scores (Situation). As the person reviewing quality, I had to sanity-check it (Task). I noticed it consistently down-ranked candidates with non-traditional career paths — a likely bias signal — so rather than just object, I pulled twenty examples showing the pattern and presented it (Action). We paused automated filtering and kept the tool only as a non-binding signal with human review (Result). My principle now: the more consequential and the more it touches people, the more a confident AI output needs human verification before it drives a decision (Reflection)."
Rubric: Strong answers show the candidate exercising independent judgment about a real AI risk (bias, hallucination, harmful edge case from Topic 7), backing the pushback with evidence, and landing on a thoughtful trust-vs-verify principle — plus the maturity to raise it constructively. Weak answers either blindly defer to the AI, reject it on vague gut feeling with no evidence, or show no awareness of the ethical/risk dimension or the stakes.