9.1 What an agent is — and when you don't need one
You can tell an agent problem from a workflow problem.
Before:08. Large Language Models
The first agent decision is whether you need one at all: when the steps are known in advance, a deterministic workflow with an LLM step is cheaper, faster and more reliable than an autonomous loop. Autonomy buys flexibility and costs predictability. It opens the agent module with its most valuable judgement. The fashion to resist is agent-washing — rebuilding what a function call already did, with more latency, more cost and a new failure surface.
Work through these
Agent vs. chain vs. plain function call
An agent decides its own next step, a chain follows a fixed sequence, and a function call does one thing. Most problems people bring to agents are actually one of the other two.
Autonomy levels and blast radius
How much a system is allowed to decide, and how much damage a wrong decision can do, are two separate dials. Setting them independently is what makes autonomy safe to grant.
Deterministic workflows with an LLM step
A fixed sequence with a model at one step gets most of the benefit with far less unpredictability. This is the design most production systems should start from.
Full Stack LLM Bootcamp · CourseThe cost of agentic loops
An agent that loops makes many model calls per task, and the cost and latency multiply accordingly. Knowing the multiplier before you build is what keeps the idea honest.
Sign in to keep your progress.
Free resources
We haven't checked most of these for screen reader use yet.
Links last checked 29 Aug 2026.
Stuck here?
Ask a mentor. A real person answers, and they can see exactly which topic you're on. Usually within a couple of working days.
Checking your session…
Topics shown in module order.