The Deployment of AI to Infer Employee Skills: Insights From Johnson & Johnson’s Digital-First Workforce Initiative
Citation. Tona, O., Leidner, D. E., van der Meulen, N., Wixom, B., Nunes, J., & Shagam, D. (2025). The deployment of AI to infer employee skills: Insights from Johnson & Johnson’s digital-first workforce initiative. Information Systems Journal, 35(6), 1516–1527. https://doi.org/10.1111/isj.12594
Summary. This open-access case study examines Johnson & Johnson’s deployment of an AI-enabled Digital Talent Platform that inferred employee skills from personal HR data to support a digital-first workforce initiative. Based on 26 interviews with 22 stakeholders and internal project documents, the authors identify three practices that supported deployment: blueprinting the future workforce, managing ethical data work across borders, and compensating for AI blind spots. The source is useful for the wiki because it connects AI-enabled skills architecture to competency definition, human-capital measurement, privacy governance, and readiness for digital transformation.
Key claims
- J&J’s deployment depended on first defining a future-oriented skills blueprint: the team identified capabilities, translated them into granular skills, and used subject-matter experts to make the language both strategically meaningful and understandable to employees at different skill levels.
- The AI skills-inference system used employee personal data, so deployment required active privacy, HR compliance, HR data governance, workers’ council, and local-regulatory work; GDPR and European workers’ council approval delayed the EMEA rollout by about 9 months.
- The organization treated AI-inferred skills as directionally accurate rather than definitive, and compensated for model blind spots by asking employees to update source data, letting employees and managers review inferred skills, and calibrating model output based on human input.
- Integration into workforce planning required aggregated leadership dashboards and routines for using inferred skills data in conversations about capability gaps, internal mobility, and digital transformation rather than treating the model as a standalone HR prediction product.
- The case’s boundary conditions matter: the practices are drawn from a large, geographically dispersed organization with a mature centralized IT unit, and may not transfer cleanly to smaller organizations where managers can assess skills directly or to AI systems using non-personal data.
Evidence & limitations
- Evidence comes from a qualitative single-case study of J&J’s Digital Talent Platform, including 26 interviews with 22 stakeholders between 2021 and 2023 and internal documents used to communicate the project.
- Two coauthors were J&J HR participants, which gives the account implementation access but also makes it especially important to treat the case as situated evidence rather than neutral proof that AI skills inference works generally.
- The paper gives rich deployment-practice evidence but does not provide independent validation evidence for the accuracy, adverse-impact properties, or downstream workforce outcomes of the inferred skills data.
- The PDF is an open-access Wiley version of record under CC BY-NC-ND 4.0; extraction was high confidence and no model-directed prompt-injection language was found.