> For the complete documentation index, see [llms.txt](https://docs.artific.nl/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.artific.nl/en/getting-started/the-three-modules.md).

# The three modules

A closer look at AI assistants, the AI Toolbox and the conversation module, plus when each is the right one.

They overlap deliberately. The same underlying capability is offered in three shapes, because the same task feels very different depending on how often you do it.

## AI assistants: for questions

**The shape:** a conversation. You type a question, it answers, you follow up.

**Best for** anything open-ended or one-off. Exploring a topic, understanding a document, drafting something unusual, asking a question you did not anticipate having.

**What makes them useful:** each one has been given a specific job and the material to do it. A well-built assistant is narrow (an HR assistant, a product support assistant), which is why the answers are better than a general chatbot's.

Depending on how it was set up, an assistant may show its sources, work with files you attach, use tools to look things up, take spoken input, or hand you over to a colleague.

→ [AI assistants for users](/en/for-everyone/assistants.md) · [Building assistants](/en/for-administrators/assistants.md)

## AI Toolbox: for repeated tasks

**The shape:** a form. Fill in a few fields, get a finished result.

**Best for** work you do the same way every time. Standard letters, summaries in a fixed format, translation with consistent terminology, rewriting into house style, extracting data from documents.

**What makes it useful:** the thinking happens once. An administrator works out the instructions, tests them, and publishes the element. Everyone else fills in three fields.

This is usually the fastest route to adoption, because it asks nothing of the user. People who would never write a good prompt will happily complete a form.

→ [AI Toolbox for users](/en/for-everyone/toolbox.md) · [Building elements](/en/for-administrators/toolbox.md)

## Conversation module: for incoming messages

**The shape:** a shared inbox. Messages arrive with a draft reply already written.

**Best for** a team mailbox that receives more than the team can comfortably answer, where most replies need the same background and tone.

**What makes it useful:** the work is reversed. Instead of reading a message and starting from nothing, you read a message *and a proposed reply*, and correct it. That is a much smaller job.

Nothing is ever sent automatically. A person always presses send.

→ [Working in the Inbox](/en/for-everyone/inbox.md) · [Setting it up](/en/for-administrators/conversations.md)

## Choosing between them

| If the task is…                                      | Use                                       |
| ---------------------------------------------------- | ----------------------------------------- |
| Different every time                                 | An assistant                              |
| The same every time                                  | A Toolbox element                         |
| Triggered by someone e-mailing you                   | The conversation module                   |
| Spread over days, with the same files and background | A [project](/en/for-everyone/projects.md) |

A useful test: **if you can write the instructions once and only the input changes, it should be a Toolbox element.** If you cannot predict what will be asked, it should be an assistant.

## How they share the same foundations

All three draw on the same underlying setup: the same [knowledge base](/en/for-administrators/knowledge-base.md), the same [tone of voice](/en/for-administrators/tone-of-voice.md), the same [tools](/en/for-administrators/tools.md), the same access model.

That is the point of a platform rather than three separate products: improve a document in the knowledge base and answers get better in all three at once.