> ## Documentation Index
> Fetch the complete documentation index at: https://dev-docs.altrum.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Run an assessment

> Answer the assistant's questions about an AI system, attach documents, pause and come back, and submit your answers.

Start from the AI system's page: choose **Start assessment**, or **Resume assessment** if one is open. The page is titled **Assess** followed by the AI system's name.

## Answer the assistant

Choose **Start the assessment**. The assistant asks about your AI system one question at a time, and its answer appears as it is written.

* Type your answer in **Type your answer** and send it. A message is up to 4000 characters.
* The assistant answers one message at a time. Wait for its reply before sending the next.
* When the assistant has summarised your answers, it asks whether you are ready. Choose **Yes, submit my answers** or **No, I want to change something**.

## Attach documents

Attach documents that describe the AI system, such as a design note, a policy or a data flow, and the assistant reads them.

| Rule | Limit |
| - | - |
| Types | pdf, doc, docx, pptx, xlsx, csv, html, txt |
| Files at a time | 10 |
| One file | 8 MB, or 5 MB for csv, html and txt |
| One upload | 8 MB in all |
| File names | At least 3 characters, each name once |

A document that breaks a rule is refused with the reason, such as "This file is too large. The limit is 8 MB for this type." A document cannot be removed from a conversation once sent.

## Pause and come back

**Save and exit** takes you back to the AI system. Your answers are kept, and the assistant carries on where you stopped. The assistant also finishes an answer it was writing when you left.

The list beside the chat holds every conversation this assessment has had, newest first. Choose an earlier one to read it: it is read only. To start the questions again, choose to start a new conversation and confirm. The current conversation is kept, read only.

If the connection drops, the page says so and offers **Send again**.

## Submit

When you choose **Yes, submit my answers**, the page shows each step:

1. Saving your answers
2. Sending them for assessment
3. Finding your risks and controls

When it is done, the results open. If a step fails, your answers stay saved and **Try submitting again** carries on from where it stopped. Nothing is submitted twice.

Submitting switches on, in your guardrail draft, the guardrails that the scope's controls map to and that were off. They start in **Log only** with nothing chosen yet. Nothing goes live until you apply it. See [Assessment results](/assessments/results).

## Messages you may see

| Message | What it means |
| - | - |
| The assistant is still answering the last message. Wait for it to finish, then send yours. | Wait for the reply. |
| The assistant is busy or cannot be reached. Your answers so far are saved: try again in a few minutes. | Nothing is lost. Try again later. |
| Altrum has not identified any risks yet. Try again in a few minutes. | Submitting is still finding risks. Choose **Try submitting again** later. |
| This AI system is still being set up. Try again in a few minutes. | The AI system was added moments ago. Try again shortly. |


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.