Common AI knowledge
ExampleThe same six-area Core-6 assessment across a cohort
Useful forWho understands foundational concepts, risks and oversight
Does not establishHow they perform a specific task
Practical guide
How do you measure workforce AI readiness?
Begin with the decision you need to make. A hiring decision, a learning plan and an AI rollout all need evidence, but they do not require the same interpretation of that evidence.
Define what “ready” means for this decision
Start by naming the people, work and AI systems involved. A recruiting team screening candidates for a role needs a fair, job-relevant decision process. An L&D team needs to find learning needs across a cohort. An AI transformation lead needs to understand both workforce knowledge and the conditions for safe adoption. Those are related questions, but no single test can answer them all.
Write down the decision before selecting a measure: what would you do differently if the result were high or low? If no action follows, a score is unlikely to be useful.
Three layers of evidence
Match the method to the question
Work in context
ExampleA separately designed interview or relevant work sample
Useful forHow knowledge is applied to a defined role or workflow
Does not establishA comparable organisation-wide knowledge baseline
Operating conditions
ExampleTools, policies, data access and manager support
Useful forWhether people have the environment to use AI safely
Does not establishIndividual understanding on its own
The genAssess product is the common Core-6 knowledge assessment delivered through Sova. The other methods above are separate activities an employer may choose to run.
A five-step measurement cycle
Identify the cohort and purpose
Decide whether you are measuring candidates, current employees, managers or another group. Make the purpose and intended use clear to participants.
Establish common knowledge
Use a consistent measure of foundational AI knowledge. Core-6 assesses six areas, including data governance, output evaluation and human oversight, in a 48-question operational form on Sova.
Add work and risk context
Consider which systems people actually use, where an error could harm others and what decisions remain human. For a specific job, assess relevant performance through a separately designed method when needed.
Choose the intervention
Turn knowledge gaps into targeted learning, clearer guardrails or further inquiry. Avoid treating a single overall score as a label that follows a person across unrelated roles.
Review change and impact
Reassess knowledge where appropriate, then check whether people can apply it in the real workflow. Monitor whether the process works fairly for the populations involved.
Example: hiring for a TA role
First measure the candidate’s Core-6 knowledge. Then, outside the genAssess product, ask a structured interview question about checking an AI-generated shortlist or run a job-relevant work sample. Consider the evidence together, with a human decision-maker.
Avoid false precision
Do not confuse one signal with readiness as a whole
Confidence surveys, usage counts and course attendance have legitimate uses, but they answer different questions from a knowledge assessment. A knowledge score still cannot prove that someone will use AI well in every task, or that an employer has suitable data, controls and leadership support.
Make an assessment consequential only to the extent supported by validation for that use and by a fair decision process. For EU AI Act literacy work, assessment may contribute evidence, but the law does not prescribe a Core-6 score or certify a particular level. Read the Article 4 explainer
The in-market assessment
See the Core-6 assessment in Sova.
One common AI knowledge assessment for candidates and employees, delivered through Sova.