AI Foundationsintermediate14 min · 11 steps
Autonomous Agents: Reason, Plan, Act, Check, Stop
How an autonomous agent reasons and plans, executes tools, learns from the environment's feedback, and stays under human control with clear stopping conditions.
iterations0
committed0€
requesterrortool calldataresponse
metrics are illustrativeProblemStep 1 of 11
Some tasks can't be written as a fixed workflow
“Book a venue for a 40-person offsite in Lisbon on May 12–13, under €8,000.” You can't hard-code the steps. Which venues exist, which are free and what they charge only becomes clear while doing the task.
A fixed chain breaks the first time a venue is taken. The system has to decide its next step from what it just learned.
← Previous lessonAgent Workflows II: Parallelize, Delegate, RefineNext lesson →Agent Design Principles: Simple, Transparent, Well-Tooled
Lesson 31 of 34 in AI Foundations