Meeting vocabulary companion

An exclusive resource for students of Arun's Introduction to Agentic AI and AI Agents course.

In a meeting about AI you rarely need to recite a definition. You need a sentence that helps people decide what the AI should do, or see why a result needs attention. Choose a sentence that fits the discussion and adapt the example to your work, keeping the meaning. Each row names the lecture where the idea was explained, so you can go back to it.

1. 01AI agent“What goal can this AI agent complete, and which steps can it choose?”Revisit it in Lecture 1.1: What Can an AI Agent Do Beyond Giving You an Answer?
2. 02agentic“The work is agentic: it chose another search because the first page did not answer the question.”Revisit it in Lecture 3.1: Is ChatGPT an AI Agent?
3. 03agent-washing“That may be agent-washing: the advertisement promises customer support, but the demonstration only shows a draft reply.”Revisit it in Lecture 3.1: Is ChatGPT an AI Agent?
4. 04model“Let's try another model with the same files and request.”Revisit it in Lecture 2.1: Why Can AI Explain a Job Without Doing It? (Models and Actions)
5. 05tools“Which tool actually creates the appointment?”Revisit it in Lecture 2.2: How Does an AI Agent Take Action? (Tools and Connectors)
6. 06connector“Does the calendar connector let us create events?”Revisit it in Lecture 2.2: How Does an AI Agent Take Action? (Tools and Connectors)
7. 07API“Can our order system accept a cancellation request through its API?”Revisit it in Lecture 2.3: What Is an API? (How Software Requests Information and Actions)
8. 08context window“The context window is nearly full. Let's save the key instructions in a file and ask the app to read them again after compaction.”Revisit it in Lecture 2.6: What Does an AI Agent Remember? (Context Window and Memory)
9. 09tokens“The token count includes pieces of the text supplied to the model and the text it generates.”Revisit it in Lecture 2.6: What Does an AI Agent Remember? (Context Window and Memory)
10. 10trigger“What is the trigger: our request, a scheduled time or an incoming message?”Revisit it in Lecture 3.3: What Starts an AI Agent? (Manual, Scheduled and Event Triggers)
11. 11workflow“Does this workflow follow prepared steps, or can the model choose what to do next?”Revisit it in Lecture 3.2: How Do Chatbots, Fixed Workflows and AI Agents Fit Together?
12. 12LLM“The LLM drafts the reply, but we still need to check the order status against the tracking update.”Revisit it in Lecture 2.7: What Is an LLM in an AI Agent?

When a seller shows you a demonstration

• Ask what the AI agent can complete and which steps it can choose. You are asking about work the seller can show.
• If someone asks why you call the work agentic, give the action and the reason: it chose another search because the first page did not answer the question.
• If the demonstration does less than the advertisement promises, name the gap: that may be agent-washing.

When the team wants AI to take an action

• Ask whether the calendar connector lets you create events. That question is about the actions a connection makes available.
• Ask which tool actually creates the appointment, to make the action more specific.
• For a different system, ask whether your order system accepts a cancellation request through its API.

When the team compares or checks AI answers

• Suggest trying another model with the same files and request, so the comparison is fair.
• When someone mentions an LLM, connect it to the job: the LLM drafts the reply, and the order status still needs checking against the tracking update.

When a long conversation runs short of space

• If the application's context display shows little space left, suggest saving the key instructions in a file and asking the app to read them again after compaction.
• If someone asks how the text is measured, explain that the token count includes the text supplied to the model and the text it generates.

When you discuss how the work runs

• Ask whether the workflow follows prepared steps or the model can choose what to do next. That question is about the work.
• Ask what the trigger is: your request, a scheduled time or an incoming message. That question is about the start.