The Future of Agentic AI

The course began with a system that predicts the next token and ended with systems that use tools, hold state, take part in human workflows, and reshape how organisations operate. Every layer was built on the one below it, and the limits of the base layer are still present at the top: plausibility is not truth, fluency is not judgement, and autonomy is a decision someone has to make.

As models become more capable and agents more deeply integrated into software and organisations, the questions in this course do not go away. They become more consequential. Where should the agent stop? Who is accountable? How do you know it is working? The systems will change. The questions are yours to keep asking.