AI Agents

KEY TAKEAWAY

An AI agent can take actions in the real world — not just answer questions, but actually do things.

💡 EVERYDAY ANALOGY — BEFORE THE TECH

🤖

An AI agent acts — a chatbot only answers

Tell a chatbot to book a flight and it gives you options. Tell an AI agent to book a flight and it goes to the booking site, searches, compares options, and books the best one — then reports back. Agents pursue goals. Chatbots respond to messages. The difference is action versus information.

👁️ Visual Explanation

🤖

AI Agent loop: Perceive → Plan → Act → Observe → Repeat until task complete

🌍 In Plain English

Traditional AI responds to your messages. AI agents pursue goals. You give them an objective and they break it into steps, use tools, check results, and keep going until the task is done.

Real-World Example

GitHub Copilot Workspace can take a bug report, write code to fix it, run the tests, check if they pass, and iterate — all without you touching the keyboard.

🎓 For Students

An AI agent could research a topic for you: search multiple sources, read them, extract key points, and write a structured summary.

💼 For Professionals

Sales teams use AI agents to research prospects, draft personalised emails, check CRM for history, and schedule follow-up reminders — autonomously.

💬 Example Prompt to Try

"You are a research agent. Find the top 3 AI startups that raised funding this week, summarise what they do, and explain why each matters."

🔧 Technical DepthOptional — expand when ready

AI agents use a loop: perceive (gather info from environment), plan (decide next action using LLM reasoning), act (call tools like web search, code execution, APIs), observe (check result), repeat. The ReAct framework (Reason + Act) is commonly used. Agents can spawn sub-agents for parallel tasks.

Key Terms

Tool callingThe ability of an LLM to invoke external functions like search, calculators or APIs

Agentic loopThe perceive-plan-act-observe cycle that an agent repeats until a goal is achieved

OrchestratorThe agent that manages other sub-agents and coordinates the overall task

⚠️ Common Misconception

AI agents are not autonomous robots. They still make mistakes, hallucinate, and need human oversight — especially for high-stakes tasks.

✅ Quick Quiz

What makes an AI agent different from a regular chatbot?

AI agents need no human oversight because they always work correctly.

🧑‍🏫 Ask the AI Tutor

✦ TRY THIS IN SUMQ — QUERY PRE-FILLED

Try asking SumQ to research a topic — it acts as a research agent across multiple AI models →

Query that will be auto-filled:

"You are a research agent. Find the top 3 AI startups that raised funding this week, summarise what they do, and explain why each matters."

Related Concepts

Prompts & Prompt Engineering

7 min

›⚙️

RAG — Retrieval-Augmented Generation

8 min

›

← Back to all concepts