What is Agentic AI?
Understanding autonomous AI systems in education
From Tools to Agents
Most of us are familiar with AI tools—spell checkers, search engines, recommendation systems. You use them when you want, they do what you ask, and they stop when you're done.
Agentic AI is different. An agent is a system that:
🎯 Has Goals
An agent pursues objectives, not just tasks. It wants to achieve something.
🔄 Acts Autonomously
It makes decisions and takes actions without constant human direction.
👁️ Perceives & Adapts
It observes its environment and adjusts its behavior to reach its goals.
⏰ Persists Over Time
It continues working toward goals across multiple interactions and contexts.
The Agent Loop
Think of an agent as constantly cycling through:
Perceive
Observe the current situation
Decide
Choose an action based on goals
Act
Take action in the environment
Learn
Update understanding and strategy
Examples in Education
Not an Agent: Grammar Checker
- Passive until you use it
- Responds to specific text
- No goals beyond highlighting errors
- Doesn't persist or follow up
Agent: AI Writing Tutor
- Goal: Improve student's writing over time
- Observes patterns in student work
- Proactively suggests exercises
- Adapts difficulty based on progress
- Initiates check-ins and interventions
Not an Agent: Search Engine
- Responds to queries
- Returns results passively
- No ongoing relationship
Agent: Research Assistant AI
- Goal: Complete research project
- Searches multiple sources autonomously
- Synthesizes findings
- Identifies gaps and pursues new leads
- Adjusts strategy based on what it finds
Why Does This Matter?
When AI moves from tool to agent, the power dynamics shift dramatically:
- Less human control: You're not directing every action—the agent is making decisions.
- Opacity: Complex agent behavior can be hard to predict or explain.
- Goal alignment: What if the agent's goals don't perfectly match educational values?
- Relationship changes: Students may form attachments or dependencies on agents.
- Accountability: When an agent makes a harmful decision, who's responsible?
In education, these questions become especially urgent because they involve children, power imbalances, and the fundamental purpose of learning.
A Humanities Perspective
From a humanities standpoint, agentic AI raises profound questions:
📖 Literature & Philosophy
What does it mean to learn from a non-human agent? How does this change the nature of mentorship, wisdom, and knowledge transfer?
🏛️ History & Politics
Who controls these agents? What historical patterns of technological adoption and inequality are we repeating?
🎭 Arts & Creativity
Can an agent nurture genuine creativity, or does it optimize for measurable outcomes at the expense of exploration?
⚖️ Ethics & Values
What values are embedded in agents? How do we ensure educational AI serves human flourishing, not just efficiency?