Turn a short set of meeting notes into an action list and you’ll meet ten AI terms in the order they become useful. Start with what does the work, then follow what you send, what comes back and what you need to check.

Example meeting notes and brief

Mina will send the revised agenda by Thursday. Someone needs to book a room. Lee will share the slides. The previous meeting ran ten minutes over.

Turn these notes into a table with one row per action and the columns Action, Owner and Due. Use only the notes. Write “Unassigned” for a missing owner and “Not stated” for a missing deadline. Keep “Thursday” as written. Leave out observations without an action.

What does the work

  1. Artificial intelligence (AI) covers computer systems that perform tasks such as recognizing patterns and processing language. Here the task is to identify commitments in ordinary sentences. Check whether the system picks out commitments as well as recognizing names.
  2. Generative AI produces content from learned patterns. Here it drafts a table from the notes. Check whether the new wording preserves the meaning of each commitment.
  3. Large language model (LLM) is the type of model doing the language work in this example. It can turn “Lee will share the slides” into a compact action row. Check names and dates against the text because plausible continuations can introduce details.
  4. Chatbot is the conversational software you type into. It collects the brief and notes and displays the table. A chatbot can also use rules or call other services, so the chat interface alone tells you little about which model handled a request.

What goes in and what comes back

  1. Prompt is the instruction or question that guides the response. “Turn these notes into a table” names the task. The rest of the brief makes omissions and missing fields checkable.
  2. Input is what you supply for processing. Here it includes both the prompt and the notes. Put the notes in a separate block so the model can distinguish source material from the instructions.
  3. Output is the result the system returns. Here it’s the action table. Check both its shape and whether each cell is supported.
  4. Context is the information available to the model for this response. The brief and notes are in context, and earlier chat turns may be included. Include a meeting date if you need to resolve “Thursday” into a calendar date.
  5. Token is a unit of text a model processes. The notes, brief and answer use tokens. A word can occupy several tokens, so use the tool’s token count when working with long transcripts. OpenAI’s explanation of token counting shows how text can be split.
  6. Hallucination is a false or unsupported detail presented in generated content. If the table assigns the room to Mina, look for evidence in the notes. The room stays unassigned when the source gives no owner.

Example action list

ActionOwnerDue
Send the revised agendaMinaThursday
Book a roomUnassignedNot stated
Share the slidesLeeNot stated

Check your understanding

  • Which part tells the chatbot to leave out observations? The prompt sets that rule. The sentence about running over belongs to the input but supplies no action.
  • Should Lee’s row inherit Thursday? The notes give Thursday only for Mina’s agenda, so Lee’s deadline stays “Not stated.”

Try the meeting-notes exercise with a fresh example. Trace each owner and deadline to the source before sharing the list.