IntermediateLesson 5 of 58 min read
Agents: when AI takes action
A course from the Academy: Advanced user: mastering AI tools
Go to the quiz ↓So far, AI has answered you. The new generation acts: these are agents, the big shift under way.
Chatbot vs agent
Ask a chatbot "find me a cheap flight to Lisbon": it gives you advice. An agent, by contrast, opens the comparison sites, compares prices, fills in the forms and comes back with concrete options, or even books, if you authorise it. The difference: the agent has tools (browser, files, calendar, code) and an autonomy loop: it plans, tries, checks the result, corrects, starts again.
What already exists
- Coding agents that write, test and fix programs on their own.
- Research agents that read dozens of sources and deliver a synthesis report.
- Assistants that sort emails, draft replies, manage the calendar.
- Standards like MCP, "universal sockets" that plug AIs into applications.
The golden rule: autonomy must be earned
Giving tools to an AI means giving it the power to take real actions, and therefore to make real mistakes: an email sent, a purchase made, a file deleted. Serious systems keep a human in the loop for anything irreversible: the agent prepares, you approve.
Key takeaways
- Agent = model + tools + autonomy loop (plan, act, verify, correct).
- This is where the whole industry is heading.
- Irreversible actions = human approval, always.
Key concepts
Complete the lesson
A 4-question quiz: 3 correct answers are enough, plus an exercise to practise. It opens in the Academy, on this same lesson.
Take the quiz and complete the lesson Free. With an account, your progress is saved and leads to the course certificate.