GenAI skills surge as traditional AI roles face rapid transformation

A new study published in Algorithms introduces a powerful framework designed to decode the evolving skill landscape. By leveraging Kolmogorov–Arnold Networks (KANs) and explainable AI (XAI) techniques, the research, titled "Future Skills in the GenAI Era: A Labor Market Classification System Using Kolmogorov–Arnold Networks and Explainable AI," provides actionable insights for educators, policymakers, and employers seeking to bridge the gap between traditional AI competencies and emerging GenAI demands.

Mapping the AI skills landscape

The study analyzes 9,357 job postings collected from leading European job platforms between January 2023 and May 2025, covering diverse sectors and roles across the technology spectrum. Job postings were categorized into traditional AI and modern, GenAI-aligned roles, creating a foundation for understanding how the labor market is evolving.

To accurately classify these roles, the researchers developed a system named KANVAS. It uses Kolmogorov–Arnold Networks, a deep learning model optimized for handling complex, high-dimensional datasets, and integrates explainability through SHAP (Shapley Additive Explanations) to ensure transparency in how classifications are made. The framework not only separates modern and traditional roles with high accuracy but also identifies the specific skills that differentiate the two.

The classification process achieved an impressive 79 percent accuracy and an area under the curve (AUC) score of approximately 0.86, outperforming baseline models such as support vector machines and random forests. These performance metrics underline the reliability of the model in analyzing the rapidly shifting dynamics of AI-related roles.

Skills defining the GenAI and traditional divide

The findings present a clear picture of a skills bifurcation in the AI-driven labor market. Roles aligned with modern, GenAI-centric technologies emphasize emerging skills such as prompt engineering, large language models (LLMs), LangChain frameworks, AI agents, and vector database operations. These competencies reflect the demand for expertise in designing, fine-tuning, and deploying advanced generative systems that power applications in industries ranging from software development and marketing to customer service and data analytics.

On the other hand, traditional AI roles remain anchored in classical machine learning and established enterprise technologies. Skills such as DevOps, statistics, Python programming, and legacy model operations continue to dominate these positions, supporting industries where predictive analytics and conventional data science pipelines are still core operational drivers.

By using explainable AI, the study identifies not only which skills define each cluster but also the relative importance of these skills in classification decisions. This level of interpretability enables decision-makers to pinpoint gaps in workforce capabilities and design targeted upskilling programs.

Implications for education, industry, and policy

The research highlights pressing implications for stakeholders navigating the GenAI transformation.

For educational institutions, the study underscores the urgency of updating curricula to align with the modern AI landscape. Traditional courses in programming and statistical modeling need to be complemented with advanced modules on prompt engineering, LLM orchestration, and generative model deployment. The authors recommend embedding these competencies into core data science and computer science programs while offering specialized electives to meet industry needs.

In the corporate sector, employers can leverage the framework to design reskilling and upskilling strategies that bridge the gap between existing talent pools and emerging role requirements. By using the explainable outputs from the KANVAS model, organizations can map employee skill profiles against market demands, identify capability gaps, and implement personalized training initiatives.

For policymakers and labor market analysts, the findings reveal a critical need to modernize occupational taxonomies such as ESCO and O*NET. Current classifications lag behind the rapid evolution of GenAI technologies, making it challenging to capture new job categories and competencies accurately. The study suggests integrating insights from real-time labor market analytics into policy frameworks to ensure that workforce development strategies remain relevant and forward-looking.

Bridging the gap between traditional and modern AI roles

The study provides an actionable roadmap for transitioning professionals from traditional AI roles into GenAI-aligned positions. By identifying the precise skill sets that differentiate the two categories, the framework empowers organizations and individuals to target specific areas for development.

For instance, a data scientist proficient in classical machine learning but lacking exposure to LLM orchestration or AI agent deployment can use these insights to pursue targeted training, accelerating their transition into high-demand GenAI roles. Similarly, organizations can use these skill maps to inform recruitment, workforce planning, and strategic investments in training infrastructure.

The researchers also highlight the importance of explainability in fostering trust and adoption. By providing transparent reasoning for classification outcomes, the system encourages stakeholders to engage with the results confidently, ensuring that AI-driven recommendations are both interpretable and actionable.

The authors acknowledge certain limitations in their work, including regional concentration on European job postings and potential noise in automated labeling processes. However, they argue that these limitations do not diminish the broader applicability of the findings. Future research, they note, should explore multi-label classification approaches to account for hybrid roles and expand analysis to other regions and industries.