Starting a Chat
Chat is the main way you work with Arketic. You ask a question in plain language, and the assistant answers — searching your documents, looking things up on the web, or carrying out an action for you.
This page walks you through everything that happens between opening Arketic and reading your first answer.
Where a Chat Can Start
You can begin a conversation from several places.
| Starting point | Where you find it | What it does |
|---|---|---|
| New Chat | Top of the sidebar | Opens a fresh, empty chat home screen |
| Assistant chip | On the chat home screen | Selects that assistant for your next message |
| Start Chat | On an assistant's card in Assistants | Opens a chat with that assistant already selected |
| Recent | Sidebar | Reopens a past conversation and continues it |
| Inside a project | The project's chat area | Starts a conversation shared with the project team |
Nothing is saved until you send. Opening the chat screen does not create a conversation — the conversation is created the moment you send your first message or attach your first file.
The Home Screen
When you open Arketic you land on the chat home screen.
What you see there:
| Element | Description |
|---|---|
| Greeting | A time-of-day greeting with your first name — Good Morning, Good Afternoon or Good Evening |
| "What would you like to explore today?" | The prompt line under the greeting |
| Assistant chips | Up to four recommended assistants, shown as clickable chips above the input |
| Input box | The large text box with the placeholder "Ask Arketic anything..." |
| Toolbar | The row of buttons inside the input box (see below) |
| Disclaimer | "Arketic AI can make mistakes, please double-check the responses!" with a link to the Privacy Policy |
Choosing an Assistant
Different assistants are set up for different jobs — one may know your HR documents, another your IT procedures. Choosing the right one usually gives you a better answer.
To pick an assistant, click one of the chips above the input. The selected chip is highlighted with a sky-blue outline. Click a different chip at any time before you send to switch.
If you do not pick one, Arketic uses a default assistant — Arketic AI when it is available.
If no assistant is selected, sending shows the message "Please select an assistant first". Pick a chip and send again.
Browse the full list, including assistants that are not shown as chips, on the Assistants page. See Browsing Assistants.
The Assistant Welcome Screen
Once an assistant is selected and the chat is still empty, its welcome screen appears:
- The assistant's avatar
- Its name
- A short description, if the assistant has one
- Starter question chips, if the assistant has them configured
Clicking a starter question fills it into the input box — it does not send it. You can edit the wording first, then send.
The Input Toolbar
The toolbar sits inside the input box, under where you type.
| Button | Tooltip | What it does |
|---|---|---|
| + | Add File | Opens a menu for attaching files or a folder, inserting a prompt, or picking an employee |
| Sliders icon | Tools | Opens the Tools menu: one-message actions such as Create Image, plus the Web Search and Document Search switches |
| Brain icon | Thinking: off (fast response) / Thinking: on (slower, shows the thinking bubble) | Switches thinking on or off — whether the assistant reasons through your question before answering |
| Microphone icon | Dictate | Records your voice and turns it into text in the input box. Nothing is sent |
| Arrow button | Send message | Sends your message; becomes a Stop generating button while an answer is being written |
The + Menu
| Option | What it does |
|---|---|
| Upload File | Opens the file picker so you can attach one or several documents, spreadsheets, images or audio files |
| Upload Folder | Lets you pick a folder and attaches the files inside it |
| Prompts | Opens the Quick Prompts window, where you can pick a ready-made prompt. Choosing one inserts its text into the input box without sending |
| Select Employee | Opens the Select Employee window so you can ask about a specific team member |
Upload File and Upload Folder appear only if your role allows you to attach files in chat.
Select Employee is shown to people whose role is Executive or Admin — it does not appear for the Member role. Once someone is selected, a chip with their name appears at the top of the input box, and the Select Employee row in the menu shows a small dot. Remove the selection with the × on the chip. The selection applies to the next message you send.
While files are uploading, the + button shows a loading animation. You can also drag files from your computer and drop them onto the input box. See Attachments for supported file types and limits.
The Tools Menu
The Tools menu has two parts.
Actions for one message sit at the top:
| Action | What it does |
|---|---|
| Create Image | The assistant draws a picture from your description |
| Create Document | The assistant writes a document and gives it to you as a file |
| Create Presentation | The assistant builds a slide deck |
| Summarize File | The assistant reads the whole attached file and summarizes it. Shown only while a file is attached |
Clicking one prepares it for your next message only: a chip with the action's name appears at the top of the input box, and the placeholder changes to a hint about what to type. Remove the chip with its × to cancel. After you send, the action switches itself off.
Search switches sit below them and stay the way you set them:
| Switch | What it does |
|---|---|
| Web Search | Lets the assistant look things up on the web |
| Document Search | Lets the assistant search your organization's documents |
Each switch turns sky-blue when it is on.
An action or a switch only appears if the assistant is allowed to use that capability, so the menu can look different from one assistant to another. If the assistant has none of them, the Tools button is not shown at all. In project chats, all of them are available. See Using AI Tools.
Thinking Mode
The brain icon is an on/off switch for thinking — there is no menu. It is grey while thinking is off, which is the default, and turns purple with a purple background when thinking is on. Hovering it shows the current state. See Thinking Mode.
Dictating a Message
Instead of typing, you can speak.
- Click the microphone button (tooltip: Dictate). The first time, your browser asks for permission to use the microphone.
- Speak. The button turns red and a timer shows how long you have been recording.
- Click the button again (tooltip: Stop recording). A spinner shows while your speech is turned into text.
- The text appears in the input box, after anything you had already typed. Check it, correct it if needed, then send.
Dictation never sends the message for you. A recording stops by itself after two minutes and is then turned into text.
Voice features depend on your browser and on your organization's installation. In a browser that cannot record, the microphone button is disabled and its tooltip reads "Voice message — not supported in this browser". If microphone access is blocked, you see "Microphone permission was not granted. You can allow it from the permission setting in your browser's address bar." If speech is not available in your installation, you see a message such as "The speech service cannot be reached right now."
Sending Your Message
Type naturally — there are no commands or special syntax to learn.
| Key | Result |
|---|---|
| Enter | Sends the message |
| Shift + Enter | Starts a new line without sending |
The send button stays disabled while the input box is empty. If your role does not allow sending messages, it stays disabled and its tooltip reads "You don't have permission to send messages".
Examples of things you can ask:
- "What's our refund policy?"
- "Create a task for the marketing team to review the Q2 report"
- "Summarize the employee handbook"
- "Show me all open tasks assigned to me"
Say what you want the answer to look like — "as a short bullet list", "as a table", "as an Excel file" — and the assistant will follow it.
What Happens After You Send
Your message appears in the conversation and the assistant starts replying.
While You Wait
A waiting indicator shows the assistant is working. Its wording changes as time passes — from "Thinking..." to "Running detailed analysis...", "Preparing comprehensive response..." and "Almost ready..." for longer answers. If the assistant has to try again, it reads "Retrying...".
If the assistant needs to look something up, a small status line appears showing what it is doing — for example "Searching documents..." then "Documents found". After a step finishes you may briefly see "Processing tool results...". See Using AI Tools.
You can stop at any time: the send button becomes a dark Stop generating button while an answer is being written. Whatever has been written so far is kept.
The Answer
| What you may see | Meaning |
|---|---|
| Formatted text | Answers come back with headings, lists, tables, links and code blocks |
| Sources | A strip under the answer. Closed, it shows the name of the first source and +N more. Click it to see N sources: each document with a relevance percentage, plus any web pages |
| Source-grounded answer | A green badge next to the timestamp, shown when the answer is well grounded in the sources. Hover the answer to see it |
| An image | A picture the assistant created, shown inside the answer |
| A download card | A file the assistant created — see Artifacts & Documents |
| Suggested follow-up questions | Clickable chips just above the input box, shown once the answer is complete; clicking one sends it straight away |
| Side panel | Opens when the assistant builds rich content such as a document, table or diagram — see Artifacts & Documents |
Arketic AI can make mistakes. Double-check important answers, and open the sources under the answer to see exactly where the information came from.
Message Actions
Hover a message to reveal its actions. On mobile they are always visible.
| Action | Available on | What it does |
|---|---|---|
| Copy | Any message | Copies the message text |
| Read aloud | Answers | Reads the answer out loud. While it is reading, the same button becomes Stop |
| Regenerate | The last answer | Writes the answer again. The new answer replaces the current one — the tooltip reads "Regenerate (replaces this answer)" |
| Continue | An answer that was cut off | Asks the assistant to carry on from where it stopped |
| Helpful / Not helpful | Answers | The thumbs-up and thumbs-down buttons record your feedback; click the same one again to remove it |
Notes:
- Read aloud takes a few seconds to start, and only one answer is read at a time. It depends on your organization's installation: where it is switched off, you see "Read-aloud is switched off in this installation."
- Helpful and Not helpful are disabled if your role does not allow giving feedback.
Taking Actions Through Chat
You can ask the assistant to do things, not just answer questions — create a task, submit a request, start a process.
Anything that only reads information runs on its own. Anything that changes or sends data first shows an approval card titled "Action Requires Approval", listing exactly what will happen. Nothing is changed until you press Approve.
See Using AI Tools.
Chatting Inside a Project
You can also chat inside a project. Project conversations are shared with the project's members rather than being personal, they are not tied to a single assistant, and both web search and document search are available in them.
See Projects Overview.
Notices You Might See
| Notice | What it means |
|---|---|
| "{n} old messages summarized" | A long conversation was condensed so it can keep going. Earlier messages stay visible |
| A note about document access | Some document features may be unavailable depending on your organization's data settings |
| "Please select an assistant first" | No assistant is selected — pick a chip and send again |
| "Still loading this conversation…" | A past conversation is taking a while to open. If it takes much longer, "This is taking longer than usual." appears with a Try Again button |
| "This conversation is still loading — send again in a moment." | You sent a message before the conversation finished opening. Wait a moment and send again |
Finding Your Conversations Again
Everything is saved automatically once you send your first message.
- Recent in the sidebar lists your latest conversations
- All Chats shows your full history, with a search box that looks through all your conversations
Opening a past conversation reloads its full message history and its sources, and the assistant keeps the context of that conversation. See Chat History.
Your personal conversations are private to you. Only conversations inside a project are shared, and only with that project's members.
Next Steps
- Thinking Mode — switch the assistant's reasoning on or off
- Using AI Tools — let the assistant act for you
- Attachments — bring your own files into a chat
- Artifacts & Documents — get documents, tables and diagrams back
- Tips & Best Practices — get better answers, faster