Chat & Assistants FAQ
Common questions about chatting with the AI, choosing assistants, thinking mode, attachments, and what the AI can and can't do on its own.
What is the difference between assistants?
Each assistant has its own knowledge access and area of focus. Arketic AI is the general-purpose assistant and is a good starting point for most questions. Others are domain-specific (for example HR, IT, or Legal) and are tuned for that kind of work.
You pick which assistant to use for a chat, so you can match the assistant to the task. See Using Assistants.
Some assistants are restricted. If you don't see one you expect, it may not be available to your role — ask your administrator.
What is thinking mode?
Thinking mode decides whether the AI reasons through your question before it answers. It is a single switch:
| Setting | When to use it |
|---|---|
| Off | The default. The answer starts right away. Good for everyday use. |
| On | The AI reasons first and shows you its reasoning. Slower, but better for complex or multi-step questions. |
You switch it with the brain icon next to the message box. It is grey while thinking is off and turns purple when thinking is on. One tap switches it — there is no menu.
See Thinking Mode.
Where did Auto, Fast, Deep and Ultra go?
Those four levels have been replaced by the single on/off switch described above. If you used Deep or Ultra before, turn thinking on; otherwise leave it off.
Can the AI make mistakes?
Yes. AI answers can be wrong or incomplete, so always verify anything important before acting on it. For complex or high-stakes questions, turn thinking on for a more careful answer.
With thinking on you'll see a collapsible Thinking bar that changes to Thought for Xs when the AI finishes. You can expand it to follow how the AI reached its answer.
Can I change how the AI responds to me?
Yes. In Settings → AI Preferences you can set your Response Style:
| Style | What it does |
|---|---|
| Concise | Shorter, to-the-point answers. |
| Detailed | Fuller answers with more explanation. |
This applies to your own chats across the product. Settings is at the bottom of the sidebar — see Account Settings.
What files can I attach to a chat?
You can attach common document and image types, including PDF, Word, Excel, PowerPoint, text, CSV, Markdown, and images. You can also attach audio files in chat.
Each file can be up to 50 MB, and one conversation can hold up to 100 files. You can pick several files at once, use Upload Folder, or drag files onto the message box. Attaching files depends on your role's permissions. For full details, see Attachments.
Files you attach in a chat stay with that conversation. They are not added to your My Documents — that's a separate personal storage area.
Can the AI read images and audio I attach?
Yes. The AI reads text found in images, and it transcribes audio to text so it can work with what was said.
Can the AI access my personal documents?
Only documents whose access level makes them available to the AI. Personal-level documents are used only in your own conversations, not anyone else's. See Access Levels.
Can I see what documents the AI used to answer?
Yes. When an answer is based on documents or web pages, a sources strip appears under it showing the first source. Click it to list every source — each document with a relevance percentage. A green Source-grounded answer badge next to the answer's time means the answer is well supported by its sources. See Using AI Tools.
How do I ask for an image, a document or a presentation?
Just ask in words, or open the Tools menu next to the message box and choose Create Image, Create Document or Create Presentation. The action applies to your next message only — a chip at the top of the message box shows it is set, and its × cancels it. Summarize File appears in the same menu once you attach a file.
Each action is shown only if the assistant you are using is allowed to do it. If one is missing, try another assistant or ask your administrator. See Using AI Tools.
Can I talk to the AI instead of typing, or have an answer read to me?
Yes, where your browser and your organization's installation support it.
- The microphone button in the message toolbar (Dictate) records your voice and puts the text in the message box. It does not send — check the text, then send it yourself.
- The Read aloud button under an answer reads that answer out loud; the same button stops it.
See Starting a Chat.
How do I find an old conversation?
Open All Chats from the sidebar and type in the Search all conversations box. The search looks at the name of each conversation and at the first message you sent in it, across your whole history. See Chat History.
Can the AI take actions without my permission?
No. When the AI wants to run an action that creates or changes data, it shows you an Action Requires Approval card with Approve and Cancel buttons. Nothing happens until you choose Approve — if you Cancel, the action is not run. See Using AI Tools.
Are my conversations private?
Regular chats are private to you. Only project conversations are shared, and only with the members of that project.
Can I delete a conversation?
There's no delete button for a saved conversation on your chats list — conversations are kept for your reference. If you start a new chat but don't send anything, that empty conversation is cleared automatically when you begin another one.
To have real conversations permanently removed, contact your administrator with a data deletion request.