Skill Assessments: An Admin's Guide

How AI-driven skill assessments work in OneRange Vero, the features you can configure, the benefits over traditional testing, and what to look for in the analytics

For administrators and people leaders · 15 min read

• What a skill assessment is in Vero — How assessments differ from courses and final quizzes
• Creating an assessment — The three-step authoring flow and the choices you make
• How the AI actually assesses skill — Question generation, scoring, and the four-question pattern
• Single-sitting and timed assessments — When and how to enforce one-shot delivery
• Assigning and what the learner sees — How assessments land in My Training
• Why AI-driven assessment is better — The honest case versus quizzes, self-rating, and manager rating
• What you'll see in analytics — Reading the dashboard, workforce analytics, and ROI metrics
• Practical playbook — How experienced admins actually roll this out

1. What a skill assessment is in Vero

How assessments differ from courses and final quizzes

A skill assessment is a stand-alone, AI-led conversation whose only job is to figure out where someone actually sits on a given skill — beginner, intermediate, advanced, or expert. It is not a course. There is no teaching, no remediation, and no second chances mid-flow. The AI asks four targeted questions per skill, listens to the answers, and produces a calibrated level per skill plus a short narrative of strengths and gaps.

You use assessments when you need to know what someone can do before you decide what to assign them, who to promote, who to staff on a project, or where the team's real gaps are. Courses build skill. Assessments measure it.

Tip: If you find yourself asking "how good is our team at X?" — that is an assessment. If you are asking "can we get our team better at X?" — that is a course.

2. Creating an assessment

The three-step authoring flow and the choices you make

From Assessments → New, the authoring flow walks you through three steps. Nothing is saved to the database until you click Confirm & Publish, so you can experiment freely.

Step 1 — Basics

• Name and description — what the learner sees in My Training
• Skills to assess — pick up to five from your company skills catalog (mix Hard and Soft as needed)
• Estimated duration — Vero suggests a sensible default based on the number of skills
• Delivery mode — choose whether the assessment can be paused or must be completed in one sitting, and set an optional time limit

Step 2 — Context (optional)

Upload reference documents — role descriptions, internal playbooks, product specs — so the AI assessor asks questions grounded in your company's reality rather than generic textbook prompts. These documents are scoped to the assessment and never leak into other assessments or courses.

Step 3 — Confirm & Publish

Vero shows you the full preview, including the questions the AI plans to ask per skill. You can regenerate, edit the description, or push live. Publishing creates v1 of the assessment; later edits to any field flag a pending version bump so you decide when to publish v2.

Tip: Single-sitting mode is best for hiring, promotion, and certification cases where you want a clean, uninterrupted signal. Leave it off for development-oriented assessments where you want learners to think between questions.

3. How the AI actually assesses skill

Question generation, scoring, and the four-question pattern

Each skill gets exactly four questions. They are not multiple choice — the learner types or speaks an answer, and the AI grades the response against a built-in proficiency rubric (beginner → expert). The four questions are deliberately graduated:

• Question 1 — surface a baseline: can the learner define the concept and use it in a basic case?
• Question 2 — apply it: walk through a realistic scenario and pick an approach
• Question 3 — handle nuance: trade-offs, edge cases, or a contrarian prompt
• Question 4 — adapt and extend: a novel situation that separates advanced from expert

The AI watches for specific signals — correct terminology, depth of reasoning, ability to acknowledge what they don't know, real examples — and produces an assessed level per skill plus a short list of strengths and gaps. The whole thing is calibrated against a 10,000-skill taxonomy so "intermediate Python" means roughly the same thing for two different learners.

Because the conversation is AI-led, there is no question bank to leak and no fixed script to memorize. Two people taking the same assessment on the same day will get questions phrased differently and follow-ups based on what they said. Cheating by sharing answers does not really work.

4. Single-sitting and timed assessments

When and how to enforce one-shot delivery

Toggling Single sitting changes the experience materially:

• Before starting, the learner sees a Ready to begin? gate that explains the rules and the time limit
• Once started, the back button is disabled and a countdown timer is visible in the header
• If the time limit is hit, Vero auto-submits and flags the result as auto-submitted in the analytics
• Closing the tab does not reset the timer — they cannot walk away and come back fresh

Tip: Pair single-sitting with a sensible time limit (rule of thumb: 5–8 minutes per skill). Too tight and you measure typing speed instead of skill. Too loose and the constraint stops doing any work.

5. Assigning and what the learner sees

How assessments land in My Training

Assessments are assigned the same way as courses — pick people or a team, set a due date, send. Learners see them in My Training, separated into Pending and Completed sections so they don't get confused with their course load. Timed and single-sitting assessments are clearly badged so nobody hits Start by accident.

After completing an assessment, learners see their assessed level per skill displayed on the Completed card so the result feels like feedback, not a black box.

6. Why AI-driven assessment is better

The honest case versus quizzes, self-rating, and manager rating

Most companies measure skill one of three ways today, and all three are broken in different ways. AI-led conversational assessment fixes the failure mode of each.

• Calibrated — every learner is graded against the same rubric, so cross-team comparisons actually mean something
• Adaptive — follow-up questions probe where the learner is weakest, so a strong opener can't carry a weak whole
• Defensible — every score comes with a written rationale you can show in a promotion or PIP conversation
• Scalable — running an assessment across 500 people costs roughly the same as running it across 5
• Hard to cheat — no question bank, no fixed answers, conversational follow-ups expose memorized responses
• Continuous — you can re-assess quarterly to track real movement instead of relying on annual reviews

7. What you'll see in analytics

Reading the dashboard, workforce analytics, and ROI metrics

Assessment results feed into three places. Each answers a different question.

Manager / admin dashboard

• The unified assignments table now shows courses and assessments side by side, with a Type column so you can scan both at once
• Each assessment row shows status (Not started / In progress / Completed), assigned date, due date, and a View action that opens the full result
• Completed assessments expose the assessed level for each skill on the View page, plus strengths and gaps the AI surfaced

Workforce analytics

• Skill heatmap — average assessed level per skill across the company or a filtered team, split by Hard and Soft
• Distribution — for any skill, what percentage of your team is beginner / intermediate / advanced / expert
• Gaps — skills where the team's average sits well below your target level, ranked by size of gap
• Movement — how the distribution has shifted over time as people retake assessments after training

ROI metrics

• Uplift — average level change from pre-training assessment to post-training reassessment on the same skill
• Time to Competency — median number of days between assignment and reaching the target level
• Gap Closure — percentage of identified gaps that have moved up at least one level since baseline
• Efficiency — uplift per training hour invested, so you can compare programs honestly

Tip: The most useful pattern is the pre/post pair: assess a skill, assign the course, reassess 30–60 days later. The delta is the cleanest ROI number you will ever get on a training program

8. Practical playbook

How experienced admins actually roll this out

• Start narrow. One assessment, three to five skills, one team. Read the results yourself before rolling wider
• Re-assess on a cadence. Quarterly is plenty for most skills; monthly for fast-moving technical work
• Tie assessments to assignments. If someone scores beginner on a target skill, auto-route them to the relevant course
• Use single-sitting for high-stakes moments only — hiring, promotion, certification. Untimed for development
• Upload context docs. A generic Python assessment is fine; a Python-at-our-company assessment is dramatically better
• Talk to learners about results. The score is a starting point for a coaching conversation, not a verdict

Tip: If a result surprises you, open the conversation transcript before second-guessing the AI. Nine times out of ten the rationale is right there in the learner's own words