Agentic AI Explained: Autonomy, Agency & Advertising

Agentic AI is a type of software that receives a goal, breaks it into steps, and uses external tools to achieve that goal without human approval for each action. For example, in advertising, it could manage a campaign brief by checking inventory, audience segments, negotiating prices, and reporting. What's interesting is the term is often used loosely. The OECD published a paper that defines an AI agent as a single system that perceives and acts with some autonomy. Agentic AI, however, refers to systems with multiple coordinated agents working together on complex objectives with minimal supervision. The paper found these terms are often used interchangeably across industries, which has consequences for liability and oversight. The OECD paper also distinguishes between autonomy and agency. Autonomy is acting without direct human involvement, while agency is the capacity to formulate goals and adapt strategy. A thermostat, for instance, has autonomy but no agency. There are four levels of action autonomy. Most advertising deployments currently sit at the "no-action autonomy" level, where a human decides, or "low-action autonomy," where a human must approve before it proceeds. The Digital Regulation Cooperation Forum in Britain also developed a similar structure with five levels of autonomy. They.