Hello everyone,
I’m Sheikh Saif Ali, a Computer Science student exploring modern Artificial Intelligence concepts. In this post, I am sharing my understanding of the paper:
“The Rise of Agentic AI: A Review of Definitions, Frameworks, and Challenges (2025)”
Tagging for feedback:
@raqeeb_26
What Is Agentic AI?
Traditional AI usually waits for a user prompt and then responds.
Agentic AI is different because it can:
Set goals Make plans Use tools Evaluate outcomes Improve actions
This means AI is moving from being a simple assistant toward becoming an autonomous system.
Goal of the Paper
The main goal of this paper is to explain:
What Agentic AI means How agentic systems work Their advantages Their challenges
The researchers show that AI systems are becoming capable of independent problem solving.
Connection with Our Course
In our AI course we studied:
Intelligent agents
Search methods
Problem solving systems
This paper extends that concept.
A normal agent:
Perceive → Decide → Act
Agentic AI:
Perceive → Plan → Use Tools → Learn → Improve
This makes the agent more advanced than traditional AI systems.
Main Components of Agentic AI
1.Planning
The agent breaks a task into smaller steps.
2. Memory
Stores previous information for better decisions.
3. Tool Usage
Can interact with:
- 1. Search engines
- 2. APIs
- 3. External software
4. Reflection
Checks its own answers and corrects mistakes.
My Personal Insight
After reading this paper manually and reviewing parts in NotebookLM, I noticed something interesting:
The future of AI is not only answering questions but taking initiative.
That was the most exciting idea for me.
Challenges Mentioned
The paper discusses:
- - Hallucination
- - Poor reliability
- - Safety concerns
- - High computing cost
These challenges must be solved before wide adoption.
Real Applications
Agentic AI can be used in:
- Medical diagnosis
- Research assistants
- Coding assistants
- Business automation
Conclusion
Agentic AI represents a major shift in artificial intelligence. It transforms AI from a passive system into an active problem solver.
This paper helped me understand how future AI systems may work independently in real-world environments.