Define the construct
More than 900 hours of research, practitioner interviews and standards mapping shaped an expert model of the knowledge behind AI readiness.
Science & validation
Science before scores.
genAssess was built to answer a hard question: can we measure the knowledge people need to work with AI in a way that is reliable, explainable and useful?
Here is the evidence so far, including its limits.900+research hours
180initial items
289pilot participants
152refined items
Development history
From expert model to operational assessment.
The work began in mid-2024, combining talent assessment, responsible AI, governance and applied workplace expertise.
Build and refine the bank
An initial 180-item bank was piloted and reviewed for clarity, difficulty and distractor performance, leaving 152 refined items.
Test the evidence
Pilot data from 289 participants was analysed for internal reliability, construct alignment and differential item functioning.
Operationalise fairly
Balanced testlets and common anchor items support varied but comparable 48-question assessment forms delivered through Sova.
Psychometric evidence
Four questions every assessment should answer.
Evidence is presented plainly so buyers, psychologists and governance teams can decide what weight to place on a score.
Does it measure consistently?
KR-20: 0.85–0.91Pilot scales reported strong internal consistency, indicating that items within the measure are working together.
Does it align with an independent measure?
r = 0.83 with GLATScores showed a strong relationship with the academic Generative AI Literacy Assessment Test.
Do items behave consistently across groups?
DIF analysis completedItems were examined for differential functioning across age, gender and ethnicity within the pilot sample.
Is the score used proportionately?
Human decision retainedThe assessment is designed as one input alongside other evidence. It does not make hiring decisions autonomously.
What the structure tells us
One overall construct. Six useful lenses.
Early factor analysis indicates a strong general AI knowledge factor. The six Core-6 scores remain valuable diagnostic lenses for interviews, learning and workforce planning.
General AI
knowledge
knowledge
AIFAUCBERDQGOIEHOC
Responsible interpretation
Strong early evidence is not the end of validation.
What is established
The pilot provides evidence of internal reliability, convergent validity and considered item-level fairness.
What continues
Criterion validation with launch clients will examine how scores relate to meaningful workplace outcomes over time.
Put the model to work