Chatbot
A conversational interface that answers messages, often with an LLM.
Focused on dialogue; it may retrieve information but usually does not take actions on its own.
AI & agents
What is the difference among an AI agent, an AI workflow and a chatbot?
A chatbot converses, a workflow runs predefined steps with LLM calls inside them, and an agent lets the model decide which steps and tools to use until the goal is met.
- Technologies
- 3
- Difference questions
- 50
What each one is
AI workflow
A fixed sequence of steps in code where some steps call an LLM.
Predictable, testable and cheap; the developer, not the model, decides the path.
AI agent
A system where an LLM chooses tools and next steps in a loop to reach a goal.
Flexible for open-ended tasks, but needs limits, evaluation and often human approval for risky actions.
At a glance
| Aspect | Chatbot | AI workflow | AI agent |
|---|---|---|---|
| Who decides the path | Mostly scripted | Developer | Model |
| Takes actions | Rarely | Yes, fixed | Yes, chosen dynamically |
| Predictability | High | High | Lower |
| Typical tools | Chat UI + LLM | Pipelines, LangGraph graphs | Tool-calling loops |
| Use for | Q&A and support chat | Known, repeatable processes | Open-ended multi-step tasks |
The 10 most-asked difference questions
Subjective questions 10
1What is the main difference between an AI agent and an AI workflow?Asked almost every time
In a workflow, the developer defines the sequence of steps in code and the LLM fills in parts of it. In an agent, the LLM decides which tools to call and what to do next in a loop until it reaches the goal, so the path is chosen at runtime.
2What is the main difference between a chatbot and an agent?Asked almost every time
A chatbot focuses on conversation: it answers messages, possibly with retrieved knowledge. An agent pursues goals by taking actions through tools, such as searching, updating records or sending emails, and can work across many steps.
3When should you use a workflow instead of an agent?Asked almost every time
When the steps are known and repeatable, such as extract, classify, route and notify. Workflows are cheaper, faster, easier to test and more predictable, which matters for compliance and reliability.
4When does an agent justify its extra complexity?Asked almost every time
When tasks are open-ended and the steps cannot be known in advance, such as research, debugging or handling varied customer requests that need different tools in different orders.
5Can a system combine all three?Asked almost every time
Yes. A chatbot front end can route simple questions to retrieval, run fixed workflows for known processes like refunds, and hand open-ended requests to an agent, all behind one conversation.
6How does predictability compare?Very often asked
Chatbots and workflows are predictable because the developer controls the path. Agents are less predictable, so they need guardrails, call limits, evaluation and sometimes human approval.
7How does cost compare?Very often asked
Workflows usually make a fixed number of model calls. Agents may loop many times, calling the model and tools repeatedly, so cost per task is higher and more variable.
8What does 'agentic' mean on a spectrum rather than a yes/no?Very often asked
Systems range from a single LLM call, through fixed workflows, routers that choose a branch, tool-using assistants, up to autonomous agents planning many steps. The more the model decides, the more agentic it is.
9What are common workflow patterns with LLMs?Very often asked
Prompt chaining, routing to specialised prompts, parallelisation of subtasks, orchestrator-worker splits and evaluator-optimiser loops, all with control flow defined in code.
10What makes an agent production-ready?Very often asked
Clear tool descriptions, limits on steps and cost, input validation, idempotent tools, human approval for risky actions, tracing, evaluation on realistic tasks and safe fallbacks when it fails.
Practise every comparison in your dashboard.
Start learning