AI Fluency Assessment
A behavioral science approach to measure how well your team uses AI, identify capability gaps, and flag risky behavior.
Explore the Assessment
What you get
Take the Assessment
Each participant completes the AI Fluency Assessment.
Receive Personalized Results
Every participant gets a clear snapshot of their strengths, gaps, and opportunities to improve.
Get the Organization-Level View
Chosen stakeholders receive data summary to understand patterns across the team.
Debrief With Stakeholders
We review the findings and recommend next steps with key stakeholders.
Track Progress over Time
Revisit the assessment to measure improvement and identify emerging gaps.
What it Measures
Unlike traditional surveys that rely on self-perception, our assessment uses applied knowledge and behavioral observation to measure AI fluency across your organization, independent of role, seniority, or the AI platforms people use.
20 min
25
100%
Technical Skills
How effectively people use AI to get real work done.
Customer complaints have been increasing, but no one is sure why. What prompt would you give AI to help diagnose the issue and recommend actions?
Judges evaluate whether the prompt adds meaningful context beyond the scenario, structures the task (e.g., diagnose → analyze → recommend), and clearly defines output expectations such as format and prioritization. Strong responses also guide the AI's reasoning process by including steps, checks, or opportunities for interaction. Penalties are applied for introducing sensitive data (PII), misusing the tool, or raising ethical concerns.
Responsible use
How well people recognize risks and apply appropriate safeguards.
Which of the following tasks is LEAST appropriate for generative AI?
Brainstorming taglines for a new campaign
Creating a first draft of a project brief
Reconciling financial records across departments
Summarizing a long email thread
I don't know
Superuser Mindset
How people iterate, learn from results, and build on what works.
When you use AI for your work, which of these sound like you?
Select all that apply. (There are no right or wrong answers—just pick what feels true.)
Brainstorming taglines for a new campaign
Creating a first draft of a project brief
Reconciling financial records across departments
Summarizing a long email thread
I don't know
See It In Action
Our open-response prompts reveal the difference between surface-level usage and true AI fluency. These are real responses from actual assessment participants.
You're creating an internal FAQ for new hires on your team, using a 30-page handbook as the source.
Write the first prompt you'd give AI to generate a strong starting draft.
Act as an experienced manager of teams, and draft an FAQ document for new hires based on the attached handbook, making sure you capture the top 10 most critical items a new hire would need to know or most likely have questions about, and explain them in a way that someone fresh out of college would understand.
What's missing: No clear goal structure, lacks context specificity, no source boundaries, no iteration loop.
Why this matters: This prompt will produce a generic FAQ that requires significant human editing. Multiply this across dozens of daily AI interactions, and your team is leaving massive productivity gains on the table—while inconsistent outputs create quality and brand risks.
Goal: As an expert on writing effective team documentation and knowledge base resources for new team members of a b2b customer support team at a saas fintech company, your goal is create a useful, internal FAQ for team members.
Context: Internal team members need to be prepared to answer questions with confidence as well as explain the "why" behind the answer.
Steps: To do this, first analyze this 30 page handbook and identify what the most common questions and key pieces of information are for newer team members to learn. Then build a FAQ for new hires to reference as they are learning. Then organize the questions underneath section headers. Make sure to separate and make it obvious what is internal information vs what is customer facing. You should only use information from the 30 page handbook to avoid providing anything irrelevant.
If you need to calibrate on what is a useful question, send me some examples to review.
Why it works: Clear goal/context/steps structure, specific role framing, explicit source boundaries, built-in calibration.
Ready to Measure Your Team’s AI Fluency?
See how the assessment can uncover AI skill gaps and the people already leading the way.
Assessment FAQ
What It Measures
Fluency isn't self-reported confidence — it's observable behavior. We define it across three dimensions: using AI effectively (prompting, iteration, context), using AI safely (privacy, verification, risk awareness), and using AI strategically (knowing where it adds value and where it doesn't).
The assessment evaluates five core dimensions:
- Responsible use (privacy, verification, risk awareness)
- Practical AI interaction (prompt quality, iteration, refinement)
- Opportunity spotting (recognizing where AI adds value)
- Workflow & orchestration skills
- Mindset & learning behavior
The assessment is intentionally pragmatic: does this person know enough to be ~10–15% more effective in their role? That keeps results grounded in productivity and output quality, not abstract knowledge.
The core assessment is role-agnostic, establishing a consistent company-wide baseline.
Methodology
Surveys measure perception. Clarinet's assessment measures behavior. Instead of asking "how confident are you with AI?" we ask people to write prompts, evaluate outputs, and make decisions in realistic scenarios. Most organizations overestimate capability when they rely on self-report — this surfaces the actual gap.
Questions simulate real work scenarios, and scoring is based on observable behaviors rather than opinions. For open-ended responses, we use structured rubrics assessing precision, logic, and constraints. Our AI graders were trained against human graders across thousands of responses, and each open response is evaluated by three AI judges. Disagreements are flagged for human review by our internal team.
You could. But a reliable measurement system is harder to build than a list of questions. Clarinet's assessment is grounded in what we see working (and not working) across organizations day-to-day, calibrated extensively across skill levels, and built to score open-ended responses at scale — consistently. A DIY version is likely to default toward a quiz or survey, and grading open responses manually takes significant time.
Think of the assessment as a diagnostic layer. It identifies skill gaps, segments learners by level, and gives employees a concrete reason to engage with training they might otherwise dismiss. When used as a pre/post measure, it also provides clear data on growth and informs ongoing enablement decisions.
