Find out who can really use the AI tools, not who says they can.
A task-based, proctored assessment inside the tools your people already use, reviewed by a human assessor. Watched, not quizzed, and levelled by role.
Self-assessment doesn't work for AI skills
Workera's 2026 AI Skills Enterprise Benchmark Report, based on 88,753 assessments across 32,422 people, found that only 11% of employees can accurately assess their own skill level. Yet 85% of L&D leaders say they're confident in the self-reported data they collect. The same report found only 13% of employees tested as "Accomplished" in agentic AI skills before any upskilling.
Usage dashboards have the same blind spot: they show who opened the tool, not who uses it well. We wrote about both problems in why self-assessment fails and usage vs proficiency.
11%
of employees rate their own skill level accurately
85%
of L&D leaders trust their self-reported skills data
13%
of employees tested Accomplished in agentic AI skills
Source: Workera, 2026
How the assessment works
The same method sits behind every track we run, whichever tool your team uses.
- 01
A real task in the real tool
Each person works inside the AI tool they use at work, on a scoped task built around their role, not a multiple-choice quiz about the tool.
- 02
Watched, not self-reported
The attempt is proctored, so the result reflects what the person did, not what they believe they can do.
- 03
Reviewed by a human assessor
An assessor reviews every attempt, pass or fail, and judges the decisions that matter: how the task was framed, what was accepted, what was checked.
- 04
Levelled by role
Foundations, Practitioner and Expert map to different responsibilities, so a new hire and a senior engineer aren't held to the same bar.
Tools with published tracks: Claude Code, Codex, GitHub Copilot, Gemini, Cursor. Using an internal tool? We build custom assessments.
405
employees certified on Claude Code across 4 organisations, including one client that came back for a second batch. See the clients or read the case studies.
Choosing an AI certification program for employees: 6 checks
Whoever you buy from, ask these six questions before you pay. They separate evidence of skill from a certificate of attendance.
1. Is it proctored?
If nobody observes the attempt, you're measuring effort and honesty, not skill.
2. Is the task realistic?
A task inside the actual tool, close to the person's real work, tells you more than questions about the tool.
3. Does a human review it?
Automated scoring handles right answers. Judgment calls, like whether a change quietly widened its scope, need a person.
4. Are there levels?
One pass mark for everyone either fails new hires or flatters senior staff. Look for levels tied to roles.
5. Can a third party verify it?
A credential should carry an ID that an employer, client or auditor can check with the issuer.
6. How often is it refreshed?
AI tools change every few months. Ask how often the syllabus and the assessment are revised.
Building the internal case for this? See how to make the business case for certification.
Frequently asked questions
What is an AI skills assessment?
A way of checking whether someone can actually use an AI tool well at work. The most reliable versions observe the person doing a realistic task in the tool, rather than asking them to rate themselves or answer questions about it.
How do you assess employees' AI skills?
Give each person a scoped task in the AI tool they use, observe the attempt, and have an assessor review the decisions they made. Self-assessments are unreliable: Workera's 2026 benchmark found only 11% of employees rate their own skill level accurately.
Which tools can you assess?
Claude Code, OpenAI Codex, GitHub Copilot, Gemini and Cursor have published tracks. We also build custom assessments for internal or proprietary AI tools.
What do we get at the end?
A result for each person, a credential with a unique ID for each pass, and a team summary you can compare against your rollout goals. We share a sample on the scoping call.
How much does it cost?
We quote per cohort, based on headcount, tools and levels, and give you a number on the scoping call.
See what your team can actually do with AI tools.
Tell us the tools and the headcount. We'll scope an assessment and show you a sample report on one call.