AI fluency is appearing in more job descriptions and interviews (you know this well). But many teams still judge it loosely: a candidate names a few tools, sounds confident, and gets marked as “good with AI.”
That tells you frustratingly little, and you're left desperate for more. The useful question is how the person works with AI: what they hand over, what they check, where the tool fails, and what they do next.
Zapier helped make the skill more concrete by publishing an AI fluency framework and using it throughout its hiring process. In March 2026, Zapier raised its hiring bar: Capable now means AI is embedded in core work, used through repeatable systems and linked to a clear improvement in quality, efficiency or another relevant outcome.
The simplified rubric below adapts Zapier’s current four-level model for a practical interview. It's not a reproduction of Zapier’s internal hiring rubric, and the right level depends on the role.
Why AI fluency is becoming a hiring criterion
Two candidates can have the same title and similar experience, but use AI very differently; that's just the beauty of human diversity. One might use it occasionally for a first draft, while the other has built it into research, analysis, writing, or operations and knows exactly where human review is needed.
The difference now matters to hiring teams
Some have started asking about AI use because they don't want to hire someone who will struggle with the team's existing workflows.
We started asking candidates in the hiring process what their use of AI is. It’s more to not have the risk of hiring someone who would be AI-averse.”
Metaview’s data shows that the topic is becoming more common. Among interviews (where Metaview’s taxonomy identified tracked topics), around 15% contained discussion classified as AI-fluency probing. In a separate measure across the broader interview set, direct discussion of AI usage rose from 0.33% in mid-2025 to 4.54% a year later.¹
The 15% and 4.54% figures use different denominators and shouldn't be compared directly. It's important you realize they show growing interest rather than a universal hiring standard.
Read Zapier’s current hiring rubric and Metaview’s discussion of AI fluency in hiring.
The four levels of AI fluency
AI fluency is the ability to use AI effectively and responsibly while remaining accountable for the result. Strong candidates can explain what improved, what went wrong, how they verified the output, and where a human still needed to make a decision.
Zapier’s current hiring framework assesses mindset, strategy, building, and accountability. The table below simplifies those ideas for interview use. Set the required level according to the work.
| Level | What it looks like | What you might hear | How to use the score |
|---|---|---|---|
| 1. Unacceptable | AI use is limited to occasional, low-impact tasks, or the candidate cannot show a repeatable workflow or meaningful improvement. | Names tools or one-off prompts but cannot explain why they used them, what improved or how the output was checked. | Probe for limited access, role constraints, valid risk concerns or lack of experience. Reluctance alone should not be treated as an automatic rejection. |
| 2. Capable | Uses AI consistently in core work through repeatable workflows, validates the output and can show a clear improvement in quality, efficiency or another relevant outcome. | Explains a recurring workflow, why a tool or model was chosen, the before-and-after result and the checks used before the work was accepted. | A suitable target where strong individual AI fluency matters but the person is not expected to redesign systems for the wider team. |
| 3. Adoptive | Orchestrates tools or builds shared systems that improve work beyond their own tasks, with clear review steps and evidence of impact. | Describes a workflow, system or process other people use, how weak output is found and how the system has been improved over time. | Strong for roles expected to raise a team’s capability, build shared workflows or lead practical adoption. |
| 4. Transformative | Re-engineers a team or function around AI, replaces legacy work where appropriate and demonstrates measurable improvement with safeguards and human accountability. | Shows a clear before-and-after at team or function level: what stopped, what changed, what capacity was freed and where humans still own the decision. | Reserve this for roles expected to redesign how a function operates. It should be rare and role-specific. |
The main distinction is how deeply AI changes the work and whether the candidate can show repeatability, impact and accountability.
What to listen for
Here's an essential tip: Candidates who only describe AI’s wins are giving you half the picture. Stronger answers include a failure, the check that caught it, and the change they made afterward.
AI is biased toward positivity. If you’re building a candidate template, you have to be prescriptive about flags.”
How to interview for AI fluency
Start with one detailed example. For roles with a higher AI-fluency bar, confirm the claim with a relevant work sample or evidence from another stage.
1. Ask for a real workflow
Ask: “Walk me through the last meaningful piece of work where you used AI, from start to finish.”
- What outcome were you trying to achieve?
- What did you give the tool or model, and why did you choose it?
- How did the workflow change the quality, speed, or scope of the work?
- What did you verify, change or reject before using the result?
Don't stop there, or infer the level from the tool name. Look for a repeatable process, a clear result, and evidence that the candidate checked the work.
2. Test their judgment
Ask: “Tell me about a time AI gave you a convincing answer that was wrong or unhelpful. How did you catch it?”
Listen for a real check: source verification, a test case, comparison with known data, peer review, or a clear reason the output did not fit the task. Strong answers also show that the candidate owned the result instead of blaming the tool.
3. Check whether the workflow is repeatable
Ask: “Have you changed the way you or your team works because of that experience?”
This shows whether the candidate can repeat and improve the workflow, rather than describing a useful one-off prompt.
4. Test where they draw the line
Ask: “Tell me about a task where you decided not to use AI, or limited how you used it. What risk were you managing?”
- Confidential or personal information
- Security or legal restrictions
- Unreliable source material
- Bias or unfair outcomes
- High-stakes decisions that require human review
- Tasks where AI would add work rather than improve it
A strong answer is not simply “I never use AI.” It explains where the risk sits, what controls are needed, and who remains accountable.
The AI fluency interview kit
The rubric, the questions that surface each level, and a scoring guide your whole panel can use this week. Start free, and run the loop on your own roles.
Start free. No credit card.
Keep one owner
Give one interviewer responsibility for AI fluency. Four people asking “Do you use AI?” creates repetition, not better evidence.
How to score it from the interview
A rubric gives interviewers a shared reference point, but the score still needs to be tied to the candidate’s evidence. If the interviewer waits two days and scores from memory, the details that separate the levels are easy to lose.
Metaview captures the interview so the candidate’s workflow, checks, mistakes, and stated outcomes are available when the interviewer completes the scorecard. Metaview Notetaker keeps interview evidence next to the candidate record, rather than leaving it scattered across notes or memory.
We can then draft the scorecard for you, aligned with the rubric the team has set. It pulls in the candidate’s own examples, so the suggested AI fluency level is tied to evidence the interviewer can review.
The interviewer checks the draft, corrects anything that is wrong, and makes the final judgment. The AI organizes the evidence; the hiring team decides the score.
Metaview Reports can then show whether the competency was assessed where expected and how scores vary across roles, teams, or interviewers.
- Check coverage: Was AI fluency actually assessed in the stages where it was required?
- Compare scoring: Are interviewers interpreting the levels similarly when the evidence is comparable?
- Review disagreements: When one interviewer sees "Adoptive" and another sees "Capable," compare the evidence and clarify the rubric.
This does not create consistency on its own, and calibration is still needed to make sure the levels are applied as intended.
Related: Metaview’s 2026 AI & Hiring Alignment Report
How to use the rubric with your team
Start with one open role, then decide whether AI fluency is materially relevant and, if it is, agree on the required level before interviews begin.
- Capable fits roles where AI should be embedded in the person’s core work through repeatable, well-checked workflows.
- Adoptive fits roles expected to build shared systems, improve team workflows, or lead practical adoption.
- Transformative should be reserved for roles expected to redesign how a team or function operates.
If AI is not relevant to the role, do not force it into the process. When it is relevant, add the questions to the interview plan, place the rubric in the scorecard, and give one interviewer responsibility for the competency.
- Define the bar before the interview: Do not decide what “good” meant after meeting the candidate.
- Score the evidence, not the confidence: A polished answer without a repeatable workflow or clear outcome should not score highly.
- Use more than one stage for a high bar: A relevant work sample can test whether the claimed workflow holds up in practice.
- Review the rubric regularly: AI fluency will not look the same across functions or six months from now.
Useful resources: interview questions, integrations, great interviewers, quality of hire, interview quality, and pricing.
The goal is not to hire the person who talks most confidently about AI. It's to see how they use it, how they check it, and whether that level matches the job.
See the rubric in your own interviews.
Record the candidate’s example, draft the scorecard against your criteria, and keep the final decision with your hiring team.
Frequently asked questions
What is an AI fluency rubric?
It is a scoring guide that defines levels of skill at working with AI. It gives interviewers a shared reference point for judging workflow, checks, impact and accountability. Teams still need calibration to apply the levels consistently.
How do you interview for AI fluency?
Ask the candidate to walk through one recent, meaningful workflow from start to finish. Follow up on the outcome, the tool choice, what went wrong, how the work was checked and where AI should not be used. For roles with a higher bar, verify the example with a relevant work sample or evidence from another stage.
What are the four levels?
Unacceptable means the candidate has not shown repeatable, meaningful AI use for the role. Capable means AI is embedded in core work with checks and clear impact. Adoptive means the candidate builds shared systems that improve work beyond their own tasks. Transformative means they have re-engineered how a team or function operates.
Should every role require the same level?
No. Some roles may not need AI fluency as a formal criterion. Where it is relevant, set the bar according to the work rather than the candidate's seniority or confidence.
How does Metaview help?
Metaview records the interview, drafts the scorecard against the team's rubric and links the draft back to what the candidate said. Interviewers review the evidence and make the final rating. Reports can then show whether the competency was covered and how scoring varies across the team.
¹ Sources: Metaview corpus of 5.5 million captured conversations (2026). The 15% figure uses tagged-topic interviews; 4.54% and 0.33% use the broader interview set. The taxonomy may miss some AI-related discussion. Zapier: AI Fluency Rubric (2 Feb 2026) and V2 hiring rubric (31 Mar 2026).
