The session will start with opening remarks from the OECD Secretary-General Mathias Cormann and the German Federal Minister of Labour and Social Affairs, Hubertus Heil. A Ministerial Panel discussion, moderated by OECD Deputy Secretary General Ulrik Knudsen, and a keynote speech from Dr. Robert Trager, Director of the Oxford Martin School, will follow. Stephanie Ifayemi, Head of Policy at Partnership on AI, will MC.

Women are 20 percentage points less likely to have used ChatGPT compared to men in the same occupations, according to a new study. Does this reflect a greater scepticism over AI, a skills gap, a confidence gap or something else entirely? And what does this mean in practice for the men and women working in occupations highly exposed to AI? This expert panel will identify the main risks and opportunities for men and women working with AI and discuss policies to ensure that men and women benefits from emerging AI advances.

Artificial Intelligence (AI) holds the promise of radically transforming economies and changing how people work and firms organise their production. To harness the opportunities brought by AI and design public policies aimed at fostering an inclusive digital transformation, it is crucial to assess AI’s potential and the current state of AI's diffusion in firms and sectors. In this context, the session will first discuss key insights from new OECD work on the topic, notably including launching the publication “The adoption of artificial intelligence in firms: new evidence for policymaking”. This session will be then dedicated to a panel discussion where leading academics and practitioners will discuss recent evidence and challenges related to AI measurement and its diffusion across firms and sectors. They will explore the role of skills, other digital technologies and intangibles, and discuss the role of public policy in fostering AI uptake and its returns.

As part of AI-WIPS, OECD is developing a set of AI capability indicators to describe what AI can and can’t do with respect to different areas of human performance. The full set of indicators will not be released until spring 2025, but this session will provide a first look at one of these indicators, which will provide a scale of AI performance related to social interaction. The session will describe the scale and how it will link to the technical literature in AI and to available measures of human social abilities and social job tasks. The session will describe the larger goals for the creation of the full set of AI Capability Indicators, with a timeline for their release, further development, and application for policy analysis.

AI incidents and hazards reported by reputable media outlets worldwide have seen a sharp increase since 2022, as tracked by the OECD AI Incident Monitor (AIM). Data on past and present AI risks and harms offers valuable evidence for policy makers to make informed decisions about AI governance. This data not only helps in understanding recurring issues but also in identifying anomalous events and early warning signals to prevent future occurrences. Reporting AI incidents in an interoperable, standardised manner has therefore emerged as a key priority in international AI governance discussions, ensuring that lessons learned from these incidents can be effectively shared and addressed globally.

Training systems are being called upon to prepare workers for the adoption of artificial intelligence (AI) in the workplace. While initial education is important, upskilling and reskilling the existing workforce is essential to help individuals and businesses adapt and prepare for a more extensive use of AI in the workplace. Policy makers face the challenge of ensuring that training is both relevant and inclusive, ranging from training to prepare AI specialists to courses that foster AI literacy among the much larger group of workers exposed to AI. Is the training available today responding to these needs? what are best strategies to better adapt its content to current and future needs?

In many countries, upper secondary graduation requires students to pass high-stakes exams. But as AI systems become more powerful, so does their potential to disrupt high stakes testing either through misuse of the technology or through transforming what’s relevant to test. Policy makers in many countries are wrestling with these issues and have arrived at a variety of solutions from banning the use of AI on exams to allowing AI’s use in controlled ways to revising the content of the tests to minimise or highlight the role AI can play in supporting human skills. This session will bring together policy makers from several countries to discuss the debates and choices in their countries about how the growing capabilities of AI may affect their high-stakes tests.

People furthest from the labour market can have large gaps in their educational attainment, qualification and work experience, and their skills can be difficult to map using “traditional” data analytics and administrative data. Furthermore, they might face additional employment challenges that the traditional administrative data often do not capture. As such, simple statistical methods to identify people who need additional support and traditional algorithms to match jobseekers with jobs by their education and job experience fail. Can AI (and big data) help overcome these challenges and identify appropriate support and job opportunities for the most vulnerable jobseekers? This session will explore the opportunities of using AI in public employment services (PES) to help the people furthest from the labour market. The panellists will discuss untapped potential use cases of AI in PES and how to unlock their potential going forward.

The economic impact of AI is intrinsically linked to the functioning of the competitive environment. For instance, market leaders with large customer bases and scales of operation might reap higher benefits from AI adoption, leading to increased concentration and market power in their hands. Additionally, the development of large AI models is highly concentrated among a few firms, raising further concerns about the dynamics and competitiveness of the AI market. This session will feature a panel discussion with leading academics and practitioners who will explore recent evidence on the link between competition and AI developments. The discussions will aim at improving the understanding of the impact of AI on market concentration and competition, and the role of AI in fostering or hindering innovation.

Investments in AI from employers are often driven by a desire to boost productivity. But evidence on the effects of AI on productivity is still emerging. Firms that use AI tend to be more productive than other firms, but this might relate to the fact that firms using AI tend to have more digital capabilities, intangibles, and skilled workers. Some evidence, focusing on specific use cases, further indicates that generative AI may boost the productivity of workers, especially the least productive ones. In addition, workers using AI say it improves their performance. This session will focus on discussing the emerging evidence at the micro-economic level on the links between the use of AI and productivity, and on the role of policymakers to enable realising the returns to AI uptake while fostering inclusive economic growth.

Beyond creating content and interacting with human users, generative AI systems are increasingly designed to function as autonomous agents capable of making decisions on a user’s behalf. The advanced foundation models powering these AI systems enable them to pursue complex goals with minimal direct supervision. These systems can plan and execute workflows, use tools, and collaborate with both other autonomous agents and humans. While this offers significant benefits, such as increased productivity by automating low-value tasks, it also poses substantial risks. These risks include labour displacement, agents executing incorrect or undesired actions, and the potential for systemic harms resulting in unintended consequences.