Skip to Content
Home └Agents: delegating work to an AI that acts
Agents: delegating work to an AI that acts

Agents: delegating work to an AI that acts

An assistant answers you; an agent looks things up, writes, sends and pays. It is a language model inside a loop with tools: it inherits everything the model does well and badly, and adds the ability to act. That is why its mistakes are no longer sentences: they are facts, and some cannot be undone. This course teaches you to decide what it is allowed to touch, how to assign it a multi-step job and how to check what it did.

Private Course
Please sign in to contact responsible
Responsible Carlos Pascual
Last Update 2026-10-06
Completion Time 4 hours 30 minutes
Members 1
Intermediate Data, Analytics and AI
  • Block 1: What an agent can do: tools and permissions
    3Lessons · 1 hr 5 min
    • From answering to acting
    • Fixed workflow or agent: when it needs to decide
    • Block 1 exam
      10 xp
  • Block 2: Briefing an agent: plan, memory and long tasks
    3Lessons · 1 hr 5 min
    • The plan before the action
    • What the agent remembers and what it writes down
    • Block 2 exam
      10 xp
  • Block 3: Reading what it has done: the trace and the report
    3Lessons · 1 hr 5 min
    • Reading a long trace without reading all of it
    • How much to review and when to stop
    • Block 3 exam
      10 xp
  • Block 4: Autonomy and security: how much rope and against what
    3Lessons · 1 hr 10 min
    • Granting autonomy in steps
    • When the data carries a hidden command
    • Block 4 exam
      10 xp