Agents · Fast-moving · Intermediate

AI Agents

Also known as: Agentic AI

Systems where a model plans a sequence of steps, calls tools, observes results and iterates towards a goal with limited human input.

What AI Agents is

An agent is a loop rather than a single answer: perceive, plan, act, observe, repeat. Tools give it real capability — search, code execution, database queries, API calls.

How it works

The model receives a goal and a tool catalogue. It emits structured tool calls, the runtime executes them and returns results, and the loop continues until a stopping condition. Production systems add step limits, budgets, retries, approval gates for risky actions and full traces.

Why it matters

Agents move AI from advice to execution, which raises both the value and the blast radius of mistakes. Reliable scoping and permissions matter more than model choice.

Common uses

→Research and report compilation
→Coding agents that edit repositories and run tests
→Customer operations with system actions
→Data pipeline and back-office automation

Strengths

✓Completes multi-step work end to end
✓Adapts when a step fails
✓Composable with existing APIs

Watch for

✓Errors compound across steps
✓Cost and latency are hard to predict
✓Security exposure through tools and injection

Continue exploring

Reasoning Model AI Memory AI Guardrails Human in the Loop AI Workflow Automation