Artificial Intelligence (AI) · Advanced level

Advanced level · Module 2 of 6 · 22 guided hours

AI Agents and Tool Use

Build systems where models call tools and take multi-step actions, with permissions, guardrails and evaluation.

Learning objectives

  • Implement tool (function) calling
  • Build simple agent loops and understand their failure modes
  • Apply permissions, guardrails and approval for agent actions

Topics

  • tool and function calling
  • tool schemas
  • agent loops
  • planning and memory
  • permissions and human approval
  • evaluating agents

Lessons

  1. Lesson 1 Tool and Function Calling

    Let a model request calls to functions you define, and execute them safely.

    tool schemafunction callingargument validationtool results

  2. Lesson 2 Agent Loops

    Let a model take several steps — calling tools, reading results, deciding next steps — towards a goal.

    agent loopstep limitsplanningfailure modes

  3. Lesson 3 Permissions and Human Approval

    Limit what agents can do and require human approval for consequential actions.

    least privilegeread vs write toolsapproval gatesaction logging

  4. Lesson 4 Evaluating Agents

    Evaluate agents on task success, safety and efficiency, not just answer quality.

    task successtrajectory reviewsafety violationscost and steps per task

Practical activity

Give the capstone assistant two tools — policy search and a referral-template filler — with strict permissions, and test it on multi-step tasks.

Expected outcomes

  • Safe, useful tool-using systems
  • Awareness of agent failure modes

Assessment

Practical: agent passes a task test set without performing unauthorised actions.

Taught live by an instructor — online or in person. Free course lessons are for self-study and do not lead to a certificate — see registration options.