Why Agentic AI (and a Progression Framework) Matters

Artificial Intelligence is evolving fast — and so are the ways it can act on our behalf. The key difference is agency: an AI system’s ability to operate with initiative. Understanding levels of agency is a practical guide for leaders, builders, and curious readers.

Pascal Bornet’s Agentic AI Progression Framework describes six levels of AI agency, from Level 0 (no agency) to Level 5 (full agency — fully autonomous goal-setting). Here is the link to the book at amazon.ca.

What can AI agents actually do? What are their capabilities? More than that, how do they “think”? How do they approach problems? To help explain AI agent capabilities, the authors developed thie SPAR framework: Sense, Plan, Act, Reflect.

What This Framework Gives You

• Clarity: A common language for comparing AI systems.
• Planning: A roadmap to assess where you are and what’s next.
• Risk Management: More agency brings bigger ethical and governance considerations.

The Six Levels at a Glance

• Level 0 — No Agency: Systems do only what they’re told, when they’re told.
• Level 1 — Assisted Agency: Assists humans; never initiates actions on its own.
• Level 2 — Semi-Automated Agency: Takes limited independent actions within strict triggers.
• Level 3 — Conditional Autonomy: Operates autonomously in constrained contexts without prompts.
• Level 4 — Broad Autonomy: Acts independently across many tasks; humans supervise.
• Level 5 — Full Agency: Sets goals and operates independently across domains.

What to Expect in This Series

• Clear definitions for each level.
• Real-world examples and use cases.
• Benefits, risks, and governance implications.
• Practical takeaways to evaluate current tools and future capabilities.

Each post will build on the last, tracing the path from basic automation to fully autonomous agents — and the increasing responsibilities that come with each step.

Agent Types, Not a Maturity Model

Think of the progression framework as a catalog of different agent types, each suited for specific needs and contexts. It isn’t a traditional maturity model where higher levels are always better. The more you climb up the framework the higher the capabilities and the higher the autonomy.

Coming Next: Level 0 — No Agency: The Foundation of All AI