AI Skills Matrix
Observable behaviors across competencies and proficiency levels.
This matrix maps the six AI competencies to observable behaviors at three proficiency levels. Use it as a reference for hiring criteria, development conversations, or self-assessment. The behaviors align with Emeri’s E1–E4 bands: foundational matches E1, developing spans E2–E3, and stronger represents E3–E4.
How to use this matrix
Each row is a competency. Each column is a proficiency level. The cells describe what someone at that level actually DOES — observable behaviors, not abstract skill claims. Look for patterns: does someone consistently demonstrate developing behaviors across most competencies? That suggests E2–E3 proficiency. Foundational across the board suggests E1.
This is a guide, not a rubric. Real assessment requires observing work, not just checking boxes. The Emeri assessment measures these behaviors through practical scenarios; this matrix helps you recognize them in daily work or interviews.
The matrix
Swipe horizontally to compare all three proficiency levels →
| Competency | Foundational (E1) | Developing (E2–E3) | Stronger (E3–E4) |
|---|---|---|---|
| Asking well | Writes basic prompts that get usable results on familiar tasks with clear instructions. | Refines prompts when initial results miss the mark; adapts phrasing for different contexts. | Structures complex multi-step requests clearly; anticipates ambiguity and provides needed constraints upfront. |
| Checking the work | Notices obvious errors or formatting issues in AI output; asks for corrections when spotted. | Verifies factual claims against sources; catches logical inconsistencies and biased framing. | Systematically validates high-stakes output; knows which claims require verification and which are safe to trust. |
| Putting it together | Uses AI for single-step tasks; relies on manual work to connect outputs into workflows. | Chains prompts or tools for multi-step tasks; builds simple repeatable processes. | Designs integrated workflows combining AI, manual review, and other tools; optimizes for reliability and speed. |
| Getting real work done | Uses AI to draft content or gather information; requires significant manual revision to finish. | Produces work that stands on its own with moderate editing; handles routine tasks independently. | Delivers polished, publication-ready work; integrates AI seamlessly into professional output without visible seams. |
| Giving good context | Provides basic task descriptions; AI often asks for clarification or produces generic results. | Includes relevant background, constraints, and examples; results feel tailored to the situation. | Anticipates what AI needs to know; provides precise context, format requirements, and success criteria without over-specifying. |
| Keeping it safe | Aware that private information should not be shared; hesitates when unsure but may not always catch risks. | Consistently redacts sensitive data; recognizes tasks where AI introduces compliance or reputational risk. | Applies judgment about when NOT to use AI; balances efficiency against privacy, accuracy, and ethical concerns in high-stakes contexts. |
Using this matrix
This matrix is an HTML reference, not a downloadable report or validated rubric. You can print the page from your browser or cite the published URL when discussing the framework.
Limitations
This matrix describes typical behaviors at each level, not absolute requirements. Real proficiency is uneven — someone might be stronger at “Checking” and weaker at “Combining.” The Emeri assessment captures that nuance through a six-competency profile; this matrix simplifies for clarity.