Agentic AI - what it is
Agentic systems engage, interact and influence rather than sit passively. AI agents focus on specific tasks with simple workflows; Agentic AI involves multiple agents running full end-to-end workflows with significant autonomy. They need infrastructure for autonomy, long-term memory and multi-step actions, plus risk models for emergent behaviours using frameworks like MAESTRO.
Agentic systems are active participants in digital environments → they engage, interact and influence rather than sit passively, demanding distinct infrastructure, risk models and governance.
| AI agents | Agentic AI |
|---|---|
| Focus on specific tasks with simple workflows → goal-driven, autonomous task performance | Involves multiple AI agents carrying out full end-to-end workflows with significant autonomy in more complex environments |
| Not new → think antivirus software and robotic vacuums | Newer → e.g., IT incident management, managing customer returns |
- Infrastructure must support autonomy, long-term memory and multi-step actions.
- Risk models → dynamic decision-making risk modelling with real-time monitoring, audit trails, explainability and human-in-the-loop/override mechanisms; must account for emergent behaviours via behavioural simulations, scenario-based modelling and multi-agent frameworks like MAESTRO.
- Governance must be dynamic, multi-layered and proactive → the three-tier guardrail framework.
99% of 1,000 surveyed enterprise AI developers were exploring or developing agents (IBM/Morning Consult), Stanford's 2025 AI Index found agents already match human capability on select tasks with speed and cost advantages, and a major airline was held liable for its chatbot's misleading policy advice → the legal cost of insufficient guardrails.
Key terms - quick answers
What is “AI agent”?
What is “Agentic AI”?
What is “MAESTRO”?
Sources and study method
This independent lesson uses active recall, spaced retrieval and scenario practice. Read the full study method.