Reskilling for AI Adoption: Predicting Workforce Readiness Using Learning Analytics
TL;DR
It is argued that reskilling should be managed on practice and application rather than completion, that readiness predictions should direct support toward employees near the threshold, and that learning analytics should be governed as employee data with developmental purpose.
Abstract
Organisations adopting artificial intelligence must reskill large parts of their workforce, and the success of reskilling is usually measured by course completion, which says little about whether employees can actually work with AI tools. This paper examines whether learning analytics, the behavioural traces of how employees engage with reskilling, can predict workforce readiness, defined as demonstrated capability to use AI in the job, and how readiness predictions should direct support. A simulation of 3,000 employees generates readiness from a process in which practice, completion, assessment, peer interaction, prior digital skill and recency all contribute, with practice and completion interacting. A gradient-boosting model on the full set of learning-analytics features attains an AUC of 0.75, against 0.68 for completion and assessment scores alone. Practice sessions (0.09) and module completion (0.08) are the strongest predictors, with assessment scores (0.04) and prior digital skill (0.04) secondary. A targeting simulation shows that directing a fixed support budget to employees with marginal predicted readiness raises overall readiness by 4.5 points, against 3.1 points for random allocation. The paper argues that reskilling should be managed on practice and application rather than completion, that readiness predictions should direct support toward employees near the threshold, and that learning analytics should be governed as employee data with developmental purpose.
