A structured lesson workspace with readable content, hands-on examples, and a clean path to completion.
The term 'agent' is thrown around carelessly. Let us establish a strict taxonomy. An automated workflow is fully deterministic. It follows a script. An automated AI workflow uses an LLM for a specific task, like classifying an email, but the overall process remains predictable.
What makes an agent an agent
This video requires an access code. Enter your code below to unlock it instantly.
Don't have a code? Contact us to get access.
| Architecture | Path to Goal | Predictability |
|---|---|---|
| Standard Automation | Fixed sequence | 100% Deterministic |
| AI Workflow | LLM assists a step | High (Predictable I/O) |
| AI Agent | LLM decides the steps | Low (Probabilistic) |
An agent workflow has a goal, but you do not define the exact path to reach it. You provide instructions and tools. The agent reasons through the process and chooses the next action. The result is probabilistic.
An agent is an LLM that runs tools in a loop to achieve a goal. - Simon Willison
Agents fail when the goal is unclear or when the steps require strict, irreversible ordering. If a mistake cannot be corrected, do not use an agent. Use a standard workflow. Keep your tool count under 10 per agent to prevent decision paralysis.
If your process requires strict sequential execution with zero tolerance for deviation, an agent is the wrong tool for the job.
Finish the lesson once you have worked through the material. This awards ★ 30 XP.