02 · Vocabulary

 Three things people call “AI.” 

 Chatbot, workflow, agent — the words get used interchangeably in industry, and the confusion is expensive. The real distinction is a single question: who owns the control flow? 

 Definitions

 Precise, with one example each

 Chatbot Responds
 Generates a reply to a message. Control flow: one turn, then back to the human.
 Example: a support FAQ bot that answers “What’s your return policy?”

 Workflow Follows a path
 LLM calls arranged along a predefined code path — steps and routing written by a developer.
 Example: classify ticket → look up account → draft reply from template.

 Agent Directs its own process
 The LLM owns the control flow : it decides the next action from observations and state, acts, and continues until a stop condition.
 Example: “find and book the best venue” — which searches, calls, and comparisons happen is decided at runtime.

 Chatbot Workflow Agent 

 Who owns control flow Human (per turn) Developer (in code) The LLM, at runtime 

 Path across runs One turn Same steps every time Changes with observations 

 Predictability High High — auditable Lower — needs eval + guardrails 

 Best when Answering is enough Steps are known in advance The path can’t be scripted 

 Anthropic’s framing (this week’s reading): workflows use predefined code paths; agents direct their own process — and you should look for the “simplest thing that works.” (Anthropic, Building Effective Agents , 2024.)

 Two more words you’ll hear in the wild: “Agentic systems” is Anthropic’s umbrella term covering both workflows and agents (from this week’s reading). “Agentic workflow” is Andrew Ng’s popular term for LLM pipelines with reflection and tool-use — in our vocabulary, usually a workflow with agent-like moments inside its steps. Neither term settles anything by itself: when someone says “we built an agent,” ask our question — who owns the control flow? — before believing the noun.

 Most “agents” are these

 The five workflow patterns

 Click each pattern. Deterministic, auditable, cheap — and often all you need. Knowing them keeps you from building an agent where a workflow would do.

 Interactive

 Autonomy is a dial, not a switch

 Drag the dial. Value rises with autonomy — and so do risk and the oversight you owe it. The engineering discipline: pick the lowest autonomy that solves the problem. 

 Rules, no LLM Single call Workflow Agent Fully autonomous 

 Value potential 

 Risk & oversight burden 

 When NOT to build an agent: the task is deterministic; a single call is enough; or the cost, latency, or risk of runtime decisions outweighs the value. Most business problems do not need an agent — saying so out loud is a professional skill.