The use of agentic AI agents was the subject of a session at the IAPP conference in London in February, chaired by Monika Tomczak, Chief Privacy Officer at the Prosus Group.

Nils Hullen, Programme Director Global AI, Privacy and Data Regulation at IBM explained how agentic AI is different from generative AI: it has autonomy, goal orientation, and can use different tools. Agents also have awareness, and persistent memory. The first three points are especially important.

Agents are driven by an objective – for example, repair a security patch by Monday am. So there is a clear mission. The agent then breaks goals down to subtasks and keeps going until there is a result. They use external software and the internet as tools. Also, there is more awareness – a broader knowledge base. Agents can adjust their behaviour if they see a conflict arising.