Future of work

Key takeaways

• The share of firms using AI rose from about 7% in 2021 to 20% in 2025. However, adoption rates vary, with smaller firms facing cost challenges, infrastructure gaps, and skill shortages.
• Exposure to AI is not the same as automation risk. High-skill occupations are often more exposed to AI, while jobs built around routine manual or cognitive tasks may face greater automation risk.
• Workers need a mix of literacy, numeracy, digital, and complementary skills, such as critical thinking, creativity, communication, and collaboration. Specialized AI skills remain rare, at about 1% of the workforce.
• In a comparison of 10 countries in the Organisation for Economic Co-operation and Development (OECD), about one-third of job vacancies were in occupations with high AI exposure. These roles often required management, business, social, emotional, and digital skills.
• The paper calls for connected policies that strengthen education, lifelong learning, employer-led training, AI literacy, LMI, and tailored support for workers facing displacement or barriers to training.

Plain language summary

The OECD describes AI as a technology that can change many parts of the economy. Firms are using it to automate routine tasks, improve decision-making, allocate resources, and change how goods and services are produced. Adoption is growing, but remains uneven across firms, sectors, and countries. Larger and newer firms often adopt AI more quickly, while smaller firms can struggle with costs, infrastructure, and shortages of workers with the right skills.

The paper explains that AI can affect employment in several ways at once: It can automate existing tasks, create new tasks and occupations, and improve productivity. The balance among these effects will shape employment outcomes. AI may replace some routine and repetitive work, but it can also complement workers’ skills by taking on the more routine parts of a job, allowing people to focus on other tasks.

The report separates exposure to AI from the risk of automation. Managers, professionals, engineers, and other workers with advanced skills are often highly exposed to AI because it can support many of their tasks. Yet their jobs also rely on non-routine thinking and social skills that are harder to automate. Workers with lower skills in jobs characterized by predictable manual or cognitive tasks may face a higher risk of automation even when their exposure to new AI tools is lower.

The OECD argues that workers need more than specialized technical training. Literacy, numeracy, and basic digital skills provide a foundation. Communication, critical thinking, creativity, teamwork, and the ability to keep learning can help people use new tools and adapt as tasks change.

The paper includes Canada in several cross-country comparisons and profiles the AI4Good Lab, a Canadian program that combines technical and ethical AI training with mentorship for women and people from groups that face barriers in the technology sector.

Definitions

AI exposure: The degree to which AI could affect the tasks in an occupation. High exposure means AI may change or support more tasks. It does not mean that a job will disappear.

Automation risk: The likelihood that technology could perform an occupation’s tasks. This measure is different from exposure to new AI tools.

AI literacy: The knowledge and skills needed to assess AI critically, communicate and collaborate with it, and use it as a tool online, at home, and at work.

Complementary skills: Skills such as communication, collaboration, critical thinking, creativity, autonomy, and problem-solving that help people work effectively with technology and adapt to change.

Why the results matter

For the public sector

The paper connects skills policy with economic, education, and social policy. Its findings support coordinated approaches that combine current LMI with education and training, career guidance, employer partnerships, and income and re-employment support for displaced workers. The report also shows why access and equity need to be considered when governments design AI and workforce initiatives.

For employers and employment services

The distinction between tasks and occupations can help employers assess where AI may support work, where it may replace routine tasks, and where training is needed. Employment services can use the same distinction to avoid viewing AI exposure as indicative of certain job loss. Both groups can focus on the full skill mix required for transitions, including technical, management, communication, and problem-solving skills.

For post-secondary institutions and students

The paper supports curricula that combine foundational, digital, and complementary skills with specialized AI education where needed. It also points to shorter, modular, and flexible learning options for adults. Students can use this broader view of skills to prepare for changing tasks without assuming that every AI-exposed occupation requires advanced programming or data science.

For educators, career development professionals, and job seekers

Career conversations can separate a job’s exposure to AI from its risk of automation. The research suggests looking at the tasks a person performs, the skills employers request, and the parts of a role that may be complemented or changed by AI. Compared to broad predictions about which occupations will grow or decline, this approach can better support specific learning plans.

For people working with underrepresented and underserved groups

The paper finds that older workers, migrants, and people with lower formal qualifications tend to participate less in adult training. It also notes that women remain underrepresented in AI research. Flexible schedules, simpler learning pathways, career guidance, recognition of prior learning, and targeted financial and non-financial support may help address barriers related to cost, time, access, and confidence.

For people in the LMI ecosystem

The report shows the value of presenting separate measures for AI adoption, task exposure, automation risk, job-posting demand, and training participation. Keeping these concepts distinct can help people interpret labour market change more accurately. Current, detailed LMI can also help education providers, employers, and governments align training with changing demand.

For most Canadians

Most workers will not need advanced AI engineering skills. The report suggests that broad AI literacy, strong foundational skills, and complementary skills will matter across many occupations. It also suggests that learning will continue throughout people’s working lives as tools and tasks change.

Notes and important disclaimers

• This OECD paper is a cross-country policy synthesis that draws on sources with different years, measures, and country coverage.
• It does not estimate net job gains or losses in Canada. Its exposure measures do not predict whether a job will be automated.
• The Canadian case study is illustrative. It is not an evaluation of national outcomes.