The OECD’s AI and the Future of Skills (AIFS) team supports policymakers by tracking what AI systems can and cannot do across key skill domains, to better understand implications for education, training, and the workplace. One key output is the OECD AI Capability Indicators (released June 2025): a framework tracking AI advances across nine capability domains (including Social interaction, Creativity, Robot intelligence, and others), each described on a five-level scale of AI performance.
In 2026, the project’s work will focus on two priorities: keeping the indicators up to date based on the latest public evidence, and linking AI capabilities to job tasks and skills to support foresight on which roles may change, how, and what education and training pathways are most affected.
The team welcomes exchanges with Apply AI Alliance members, including on:
- Evidence sources (benchmarks, evaluations, research syntheses, sector reports) relevant to monitoring AI capabilities
- Connections with initiatives working on AI adoption and skills in strategic sectors
- Collaboration opportunities, including pilots and feedback on capability-to-task/skill mapping use cases
Contact: stuart.elliott@oecd.org
Project links:
- Logige sisse kommentaaride postitamiseks
Kommentaarid
Vastus kommentaarile An important implication of… kasutajalt Syamantak Saha
Thank you for bringing the health dimension into the discussion — it is indeed one of the most consequential intersections of AI deployment.
In addition to mapping what AI systems can currently do, it may also be valuable to consider the relative speed at which capabilities evolve compared to the pace at which healthcare and education institutions can adapt.
Many public systems operate on multi-year reform cycles, while AI capability development can shift meaningfully within months.
The resulting temporal gap may become a structural stress factor in itself.
Complementing capability indicators with some modelling of this divergence rate could help policymakers anticipate pressure points before they fully materialise.
An important implication of AI is at its intersection with Health and Medicine. The advances of AI does provide opportunity to bring latest medicine, however, the legacy methods of delivering effective healthcare has to be re-evaluated, including the legislations, training methods and delivery models that have been used in the last century.
Thank you for this important initiative.
Tracking what AI systems can do across capability domains is essential for foresight and policy design.
As capability-to-task/skill mapping evolves, there may be value in complementing this framework with an additional layer:
How human capability evolves under sustained AI mediation.
Beyond identifying which tasks AI can perform, policymakers may also need to monitor whether long-term AI integration:
• preserves or alters executive decision-making capacity
• affects skill retention in partially automated workflows
• changes autonomy thresholds in complex roles
• modifies cognitive load distribution across teams
This complementary perspective could strengthen foresight by tracking not only AI capability growth, but also human capability stability within AI-augmented environments.
I would be very interested in contributing to pilot discussions or evidence mapping around this dimension.