Chat and assistants

Command's AI assistants answer questions about security controls, frameworks and policies, and help you draft text. This page explains how to use the Chat page, which assistants there are and what each one can and cannot do, and the AI help built into other pages.

Who uses it

Admin, Security manager, Policy editor, Incident responder and Auditor can use Chat. Which assistants you see depends on your role:

Assistant	Admin	Security manager	Policy editor	Incident responder	Auditor
CISO Concierge	Yes	Yes	Yes	Yes	Yes
CISO Audit Assistant	Yes	Yes	Yes		Yes
CISO Policy Author	Yes	Yes	Yes
CISO Policy Co-pilot	Yes	Yes	Yes

Your chats are your own: no other user can see them.

The assistants

All assistants answer from sources: the shared compliance library of standards and frameworks, and the web. They cite where an answer comes from, and they say so when they cannot support an answer rather than guess. Questions outside security and compliance are declined. When your workspace language is Dutch, they answer in Dutch.

Assistant	What it does	Can it change anything?
CISO Concierge	Answers questions about security controls, evidence and frameworks. Selected by default.	No
CISO Audit Assistant	Helps you judge whether controls are adequately evidenced. Reports gaps instead of assuming compliance.	No
CISO Policy Author	Drafts control and policy text in the sections Purpose, Scope, Policy and Implementation, in English or Dutch. The text stays in the chat; copy it where you need it.	No
CISO Policy Co-pilot	Reviews and edits a policy document with you. Use it in the policy editor. See Work with the AI co-pilot.	Only in the policy editor, and only after you accept a proposal or click Allow

Important

Use the CISO Policy Co-pilot from the policy editor, not from the Chat page. Only the policy editor can show its proposals and ask your permission.

Ask a question

Click Chat in the top navigation.

Click New chat, or pick up an earlier chat from the list on the left.

Choose an assistant in the picker under the message box. Switching assistant starts a new conversation; Command shows Switched to followed by its name.

Type your question in the message box and press Enter. Press Shift+Enter for a new line.

While the assistant works, it shows Searching the knowledgebase…. The answer appears in the conversation.

Every answer is labelled Command CISO, whichever assistant you chose. Under the message box Command reminds you: AI agents can make mistakes. Check important info.

The Chat page with a conversation and an answer that cites its sources

Refer to a control, an asset or evidence

In the message box:

type @, or click control, to refer to a security control;

type #, or click Asset, to refer to an asset;

type ~, or click Evidence, to refer to an evidence document.

A list of up to eight matches appears. Use the arrow keys and Enter, or click, to insert one. It appears as a chip in your message.

The Attach button also inserts a reference to an evidence document. It does not upload a file.

Read the answer

Numbered markers in the answer are citations. Click one to open the cited document at the cited page. Word and Excel files open in a new tab.

Sources the answer drew on without citing are listed under Other sources; web pages under From the web.

Sometimes an assistant ends with a question and options to choose from. Pick one or more, or type your own answer, and click Send answer, or click Skip.

Give feedback

Point at an answer to see its actions:

Copy copies the answer.

Retry asks the last question again.

Helpful or Not helpful. After Not helpful, click Tell us why, describe What was wrong with this answer? and click Send. Command shows Noted — this shapes what I remember about you.

Manage your chats

The list on the left shows your Recent chats, named after their first message. Click Archived to see archived chats, and Active to go back. Hide the list for a focused view with the button next to it.

Each chat has a menu (⋮):

Export transcript downloads the chat as a Markdown file. Citations are not included.

Archive or Unarchive moves the chat between the two lists.

Delete deletes the chat at once, without asking. You cannot undo it.

AI help on other pages

Command's AI also works inside its pages. None of it saves anything without your action.

Where	What it does	Does it ask before changing anything?
Controls: AI assists	Drafts, enriches, maps and critiques a control	It writes into the editor; nothing is saved until you leave edit mode
Controls: critique comments	Suggests a fix per weak line	Yes: Fix control or Reject critique
Evidence: AI evidence auditor	Judges whether a file proves a control	It only records its own advisory verdict
Evidence: Draft this request with AI	Fills in an evidence request	It only fills the form; you click Send request
Policies: AI co-pilot	Discusses and edits a policy document	Yes: Accept or Reject for proposals, Allow or Don't for actions
Policies: policy generation	Drafts policy documents from your facts	It starts only when you click Generate
Assets: Add protecting control	Finds the controls that protect an asset	It ties controls only when you click a Bind button
Frameworks: Draft control	Drafts a control for a framework gap	It starts when you click; the result is a Draft
Dashboard › Threat intel: risk brief	Explains why a threat matters to you	No change; the assessment and advice are not AI

How Command protects you

Answers must be based on sources, and the assistants say when something is not supported.

Content from web pages is treated as untrusted, and some addresses are refused.

Hidden instructions in what you paste are removed before the assistant reads them.

Only the policy co-pilot can change data, and only after your approval.

Troubleshooting

What you see	What it means	What to do
The picker keeps saying Loading agent…, and sending answers Chat is unavailable — no agent is configured for this workspace.	No assistant is available to your role	Ask your Admin to check your role
No answer could be produced for this question. Try rephrasing it, or narrowing it to a specific control, framework or document.	The assistant could not find an answer	Rephrase, or name a specific control, framework or document
The agent could not answer: followed by a reason	Something went wrong while answering	Click Retry, or start a new chat
Could not update chat or Could not delete chat	Archiving or deleting did not complete	Try again

← PreviousIncidents Next →Audit log