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

# Assessments

> Understand what an assessment of an AI system does, who can run one, and how it leads to live guardrails.

An assessment finds out what could go wrong with an AI system and what should be in place to stop it. Instead of a form, you have a conversation with Altrum's assistant. Your answers fill in Altrum's questionnaire, and submitting it produces a scope: the risks that apply, the controls that answer them, and the guardrails those controls call for.

You do not need an assessment to set up guardrails, but for an AI system rated High or Very high, going live needs a recent one. See [Lifecycle](/ai-systems/lifecycle.md#the-go-live-checks).

## From conversation to live guardrails

**1. Answer the assistant**

It asks one question at a time, and reads any documents you attach. See [Run an assessment](/assessments/run.md).

**2. Submit**

Altrum finds the risks and controls, and switches on the guardrails in scope in your guardrail draft, in **Log only**.

**3. Review the scope**

Check the risks, add any the assessment missed, and see which guardrails still need your choices. See [Assessment results](/assessments/results.md).

**4. Set up the guardrails**

Add the assistant's suggested settings, finish the rest and test. See [Guardrail suggestions](/guardrails/suggestions.md).

**5. Complete**

**Complete Assessment** makes your draft the live guardrail version and closes the assessment.

## Who can run one

Owners, admins, Governance Leads and the AI system's owner can start an assessment, answer the assistant, change the risks and complete it. Everyone in your organization can read the results.

One person works on an assessment at a time. While someone else has it open, you see their name and can read but not change it. If they leave without closing it, it frees up after a few minutes.

## Statuses

| Status | Meaning |
| - | - |
| **In progress** | The conversation is under way. |
| **In review** | The answers are submitted, and you are reviewing the scope and setting up guardrails. |
| **Completed** | The guardrails went live and the assessment closed. |

An AI system has at most one open assessment at a time.

## Assess again

When the AI system changes, or its rating asks for it, choose **Start new assessment** on its page. Earlier assessments stay on its **History** tab with their results and conversations. How often each rating asks for one is in [Posture](/ai-systems/posture.md#reassessment).


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