AI Fluency Interview Questions and Scoring Rubric: A Practical Guide for Candidate Screening
TL;DR: Test AI fluency before the interview with a hands-on task using an AI assistant. Use the interview to score judgement: why they made the choices you saw. Score five dimensions against the level the role needs: Aware, Functional or Advanced. The questions and rubric below are Bryq's practical guidance for structured interviews, not research findings.
By 2024, most knowledge workers were already using AI. In that year's Work Trend Index, Microsoft and LinkedIn wrote that "75% of knowledge workers use AI at work today". So asking "do you use AI?" tells you little. The useful question is how well they use it, and in our view an interview alone won't show you that.
For the wider picture, see how to assess AI fluency in hiring and how to assess AI skills in hiring. This guide is the hands-on part.
What to test before the interview, and what to score during it
Test behaviour before the interview. Score reasoning during it. A candidate can describe great prompting habits and still accept a wrong AI answer without blinking. You see that most clearly when they do the work.
Test before the interview (observed work) | Score during the interview (explained judgement) |
|---|---|
What they hand to AI and what they keep for themselves | Why they drew that line, and when they'd draw it differently |
How they write a prompt and recover when the first answer misses | How they decide a prompt is good enough to stop iterating |
Whether they catch errors, invented facts and gaps in the output | How they'd check AI output in your context, with your data |
Whether the finished output is something you'd actually send | How they'd fit AI into your team's workflow without creating rework |
Whether they paste sensitive details into the assistant | How they think about privacy, IP and disclosure when stakes rise |
The left column needs a task; the right needs a person. Swap them and you risk rewarding how well people talk about AI over how well they use it.
Bryq's AI Fluency Assessment is one example of this approach: a simulation-based assessment of about 15 minutes in which the candidate works with a live AI assistant, writing prompts and reviewing and correcting AI output.
No assessment tool yet? A work sample can show you some of the same behaviour. Give candidates a realistic task from the role, let them use an AI assistant, and ask for the final output plus the conversation. Score both with the rubric below.
Clear criteria: what good AI fluency looks like
Bryq measures AI fluency across five dimensions, each answering one question.
Dimension | The question it answers |
|---|---|
AI Task Strategy | Can they decide what to delegate to AI vs. retain for human judgement? |
Prompting & Interaction | Can they write effective prompts and iterate when AI misses? |
Critical Evaluation | Can they spot errors, hallucinations, and gaps in AI output? |
Ethical & Responsible Use | Do they understand privacy, bias, IP, and regulatory risks? |
Workflow Integration | Can they turn AI output into professional, shippable work? |
Three rules keep the criteria matched to the role:
Set the target level first. Decide whether the job needs Aware, Functional or Advanced before you see a single candidate.
Stay tool-agnostic. Score the skill, not familiarity with one product. Tools change often. Judgement carries over.
Treat it as one signal. Bryq doesn't claim AI fluency predicts job success on its own. It's one signal alongside the other things the role needs.
AI fluency interview questions, grouped by dimension
Ask two questions per dimension and anchor each answer to something the candidate did. These ten questions are Bryq's practical guidance for a structured interview. For format and note-taking, see our guide to structured interviews.
AI Task Strategy
Question | A strong answer shows | Red flag |
|---|---|---|
1. Walk me through a recent task. Which parts did you give to AI, and which did you keep? | A clear reason for each split, tied to risk, accuracy or context | "I use it for everything" or "I never use it for real work" |
2. Tell me about a time you decided AI was the wrong tool. | A specific moment where judgement or accountability mattered more than speed | No example, or a decision made only because the tool failed |
Prompting & Interaction
Question | A strong answer shows | Red flag |
|---|---|---|
3. Your first prompt gets a vague, generic answer. What do you change? | Adds context, constraints, examples or a format; breaks the task into steps | Repeats the same prompt, or gives up and writes it by hand |
4. How do you know when to stop iterating and use what you have? | A quality bar defined before starting, and a sense of diminishing returns | Stops at the first answer that sounds plausible |
Critical Evaluation
Question | A strong answer shows | Red flag |
|---|---|---|
5. Describe an AI answer that looked right but wasn't. How did you catch it? | A concrete check: a source, a calculation, a colleague, their own domain knowledge | "It's usually accurate, so I don't check much" |
6. What do you check first when AI gives you numbers, names or quotes? | Treats specifics as claims to verify, and knows which ones invite errors | Trusts figures because they're precise |
Ethical & Responsible Use
Question | A strong answer shows | Red flag |
|---|---|---|
7. What would you never paste into an AI assistant at work? | Names categories (client data, personal details, confidential plans) and the reason | Hasn't thought about it, or "it's fine if the tool is approved" |
8. When do you tell a colleague or client that AI helped produce something? | A view on disclosure and ownership, linked to the stakes of the work | Sees disclosure as irrelevant, or is unsure who owns the output |
Workflow Integration
Question | A strong answer shows | Red flag |
|---|---|---|
9. Take me from an AI draft to something you'd send under your name. | Editing for accuracy, tone and audience | Light formatting, then send |
10. How would you bring AI into a process your team already runs? | Starts small, measures rework, keeps people informed | Changes everything at once, or hides their AI use from the team |
One weak answer is a development gap, not a verdict.
Want to see a scored profile? Book a demo and watch the simulation run.
A simple AI fluency scoring rubric
Score each dimension against observed behaviour, then compare the result with the level the role needs. The levels are not pass or fail grades. This rubric is Bryq's practical guidance; adapt the wording to your roles. It is not the scoring method inside Bryq's AI Fluency Assessment. The logic of BARS applies here too.
Dimension | Aware | Functional | Advanced |
|---|---|---|---|
AI Task Strategy | Uses AI for simple, low-risk tasks and knows to avoid it for sensitive ones | Splits tasks deliberately between AI and own judgement, with clear reasons | Designs who-does-what across a whole workflow and explains the trade-offs to others |
Prompting & Interaction | Writes clear single requests; rephrases when the answer misses | Adds context, constraints and examples; iterates with purpose | Structures multi-step work and adjusts approach as the task shifts |
Critical Evaluation | Notices obvious errors and asks before relying on output | Checks facts and figures routinely and fixes what's wrong | Anticipates where output will fail and builds checks in before it does |
Ethical & Responsible Use | Follows the rules on what not to share | Weighs privacy and IP in everyday decisions without being prompted | Sets norms for others and guides responsible adoption |
Workflow Integration | Uses AI output as a starting point with heavy editing | Turns AI drafts into finished, professional work | Builds AI into team processes and helps others do the same |
How to use it in practice:
Agree the target level per dimension before screening starts.
Score the pre-interview task first, using the "test before the interview" column of the split.
Use the interview to confirm or adjust each score, noting why.
Have two reviewers score independently, then compare.
Where Bryq fits
In hiring, Bryq's AI Fluency Assessment fits the "before the interview" side: it measures how people actually work with AI in a simulation. Each of the five dimensions is scored 0 to 100, and results map to three fluency levels matched to role requirements. No pass or fail.
Every score comes with the full chat transcript plus visible AI-reviewer notes, which you can bring into the interview: "here's where you accepted that figure; talk me through it." Separately, the Bryq platform generates a personalised interview guide for each candidate.
It's tool-agnostic, uses the same instrument for candidates and current employees, and sits in one profile alongside cognitive ability and personality traits. It's built on six peer-reviewed research frameworks, including UNESCO, SFIA and OECD. It's included in your Bryq plan. See the AI Fluency Assessment overview and the AI fluency resource hub.
Bring the evidence into your interviews. Book an AI fluency demo and we'll walk you through the simulation, the transcript and the reviewer notes behind each score.
FAQ
What are good AI fluency interview questions?
In Bryq's practical guidance, good questions ask candidates to explain decisions they actually made: what they gave to AI, how they fixed a weak answer, how they caught an error, what they'd never share, and how they finished the work. Anchor each to a real task, ideally one completed before the interview.
Should you test AI fluency before or during the interview?
In our view, both, for different things. Test how candidates work with AI before the interview, with a hands-on task. Use the interview to score their reasoning about what you observed. An interview hears how people describe their habits. A task shows them.
What levels should an AI fluency rubric use?
Bryq uses three fluency levels matched to role requirements: Aware (admin, entry-level, support), Functional (marketing, finance, ops, HR) and Advanced (strategy, leadership, AI-adjacent). Set the target level per role first.
Author
Veronika is an Industrial-Organizational Psychologist and CIPD member who works at the intersection of psychometric science and applied AI, writing on assessment design, validation, and how organizational culture shapes performance.
