AI agents move from demonstrations into organisational practice

The OECD’s 16 September paper examines how organisations are developing, deploying and governing agentic AI. It draws on practitioner interviews across multiple sectors and regions, adding accounts of implementation to the discussion of what AI agents could eventually do.

The paper covers emerging applications, benefits, obstacles and governance arrangements. Its evidence is qualitative: it offers examples of organisational experience rather than a representative estimate of adoption or a universal productivity result.

For teams considering deployment, that distinction matters. A successful demonstration answers a question about capability. An operational decision also requires an account of the permissions an agent receives, the work it can complete reliably and the person or function responsible when something goes wrong.

Continue with related research →