The AI Fluency Index (shown in the product as AI Proficiency) measures how well your employees prompt AI agents, turning prompt quality into a single trackable score so you can see who is getting value from AI and who needs help.
What it is for
Enterprise teams deploying AI agents usually have no visibility into whether their employees actually know how to prompt well. Weak prompting leads to longer conversations, more retries, and lower ROI on your AI investment.
The AI Fluency Index solves this. It scores each user prompt from 0 to 100 and rolls those scores up to the user, team, and department level. With it you can:
- Quantify prompt quality across your workforce with one number.
- Classify users into proficiency levels (Expert to Novice).
- Spot who needs training and on which specific skill.
- Track improvement over time.
It evaluates the quality of the user’s prompts only. It does not judge the quality of the AI agent’s responses.
How the score works
Each prompt is scored across six dimensions, then combined into one weighted score from 0 to 100.
| Dimension | Weight | What it looks at |
|---|
| Task clarity | 30% | Is the goal explicitly and precisely stated? (summarize this contract in 3 bullet points vs. help me) |
| Context provision | 25% | Does the user give relevant background such as role, constraints, or data? |
| Output specification | 20% | Does the user specify format, length, or structure? (as a table, in 2 sentences, as JSON) |
| Prompt length | 10% | Word count of the prompt. |
| Conversation efficiency | 10% | How few turns were needed to reach a resolution. Fewer turns scores higher. |
| Constraint specification | 5% | Are boundaries or exclusions given? (don't include X, only use data from Y) |
Proficiency levels
Every user is automatically classified into one of four levels based on their score.
| Level | Score range | What it means |
|---|
| Expert | 85 to 100 | Multi-part prompts, role instructions, defined output format, uses constraints |
| Practitioner | 65 to 84 | Clear task, some context, minimal formatting, occasional correction turns |
| Beginner | 40 to 64 | Task is vague, little context, no format, needs multiple correction turns |
| Novice | 0 to 39 | One-line prompts, no context, no format, high retry rate. Priority training target |
What you can do with it
1. The AI Proficiency report
A ready-made report appears automatically as the sixth item in the left sidebar, with no setup required. It respects the workspace date range filter. It is read-only, but you can duplicate it and customize your own copy.
It contains four charts:
- AI Proficiency Score: a gauge showing the overall workspace fluency score out of 100.
- User levels distribution: a pie chart breaking users down by proficiency level.
- % Users needing training: the share of users classified as Novice or low Beginner.
- Fluency attributes breakdown: a horizontal bar chart of scores across each fluency dimension, so you can spot the systemic weak point for the org.
2. Automatic user groups by level
Four system-managed user groups are created for you: Expert, Practitioner, Beginner, and Novice. Membership is computed from each user’s latest proficiency level and refreshes automatically every day. These groups are read-only (you cannot edit or delete them), and you can use them to filter any report or to target training at the right people.
3. Build your own charts
The AI fluency score is available as a metric in the chart builder, at the interaction, conversation, and user levels. You can combine it with existing filters such as date range, user group, and agent, and render it as a line, bar, or distribution chart. Proficiency level is available as a filter and group-by so you can slice any view by level. Hover tooltips explain what each level and metric means.
The feature is off by default. To enable it for your workspace, contact your account team.
The feature is most powerful for employee-facing AI agents, and can be enabled selectively per project.