What is the AI Assistant in Task Manager?

The AI Assistant in Task Manager is not a chat window bolted onto the side of the app. It can see your tasks, chats, notes and files - the same ones you work with every day. Instead of explaining the context from scratch, you can write “summarise the Recruitment task” or “what was agreed in the Marketing chat”. The Assistant retrieves the content, messages and attachments.

You open the Assistant with the views icon in the top bar, next to Calendar, Board and Notes. You'll find the same menu in the header of each of these windows. These windows take up the same space. They share a light/dark theme and a remembered layout.

You can also ask questions without opening the window. The reply field in every task and chat, and in Mail, has an “AI Assistant” button. It switches the whole editor to private question mode. The editor turns purple, and a bar above it tells you that only you will see the question. It will not be sent to participants. The answer appears in the thread with a FOR YOUR EYES ONLY badge.

The AI Assistant works in both licence plans. In the Pro plan, it additionally draws on the whole company's knowledge base, described on the "Company Knowledge" page.

What can the Assistant see?

Each conversation thread has its own set of sources. You can select several tasks, a chat, a note, an entire workspace and a calendar with upcoming due dates (7 days by default). The selection window works like the Files module. Click a row to select a source, use the › arrow to go one level deeper, and search across everything at once using the search field at the top.

Selecting a task or chat also includes files attached to it, both those in messages and those added in the Files module. In the tree, they appear as “Workspace / thread name”. Selecting a folder in Documents gives the Assistant access to files and notes in that folder and its subfolders. The Assistant reads long notes in chunks and retrieves more as needed.

A usage bar appears below the list of sources. The budget is 30 sources and 60,000 characters. If you select more, the Assistant will tell you what it left out. General sources are dropped first; tasks selected directly are kept. You only see sources you have access to. Permissions are checked when sources are saved and again when the context for a response is built. Mail is disabled as a source by default. The administrator can enable it in Settings → AI, but only for your own inboxes.

If the Assistant confirms an action it did not perform, a yellow “Correction” box appears below the response. It shows what actually happened.

Task Manager: AI Assistant window with a list of conversation threads and a reply.
Assistant window with a list of threads. Each thread has its own history and its own set of knowledge sources.

What do you gain?

You do not need to paste task details into the conversation or recap what was agreed last week. You do not have to open five different places to gather material for one answer. Ask about what your team is working on and get an answer based on information recorded as you work.

You can see what the Assistant used and what it left out. That makes it easier to assess the answer instead of guessing whether it reflects your team’s data.

If the administrator has enabled Mail as a knowledge source, you can ask about correspondence with a customer without showing your question to the thread participants. Your conversation with the Assistant is private. You can also return to your Assistant threads without worrying that anyone else, even an administrator, will read them.

What does this look like during a typical working week

A manager receives an email from a new client and does not know what the company does. If an administrator has enabled the “AI Assistant with internet access” permission, they write: “check example.com and tell me what this company does”. The Assistant needs to know the website address, but can build it from the domain in the email address.

The manager then starts a conversation thread for this client. They select the proposal task, sales chat and folder in Documents as knowledge sources. They ask for an update on the case and request a Word document with a summary. The model writes the content, while Task Manager assembles the file and adds a “Download …” button. Finally, the manager writes: “remind me about this task the day before it’s due”. The app calculates the reminder time from the task’s due date. If the due date changes, the reminder moves with it.

The Assistant names the thread based on the first question. A week later, the manager can find the conversation using the search field, which searches names and content. They can also use a filter to show only conversations where the Assistant made a change: created a task, changed its status or set a reminder.

Frequently asked questions

What happens if the administrator has not configured AI?

The administrator configures AI in Settings → AI. They provide a provider (OpenAI, Gemini, Claude or Grok), an API key and a model. If the configuration is missing, the Assistant will not respond and will explain why it is unavailable.

Can the Assistant see other people’s tasks?

No. The Assistant can access only the tasks and chats you have permission to access. The global administrator can access everything. If you lose permission to a source, the Assistant will skip it the next time you ask a question.

Can the Assistant make changes without my consent?

Sending an email always requires your confirmation and uses your personal SMTP. A reminder is created only after you click “Yes, do it”. A task requested with “create a task” is created only after you click “Save” in the open “Add task” window.

Which files can the Assistant read?

The Assistant can read content in TXT, CSV, MD, JSON, XML, HTML, DOCX, XLSX/XLS and PDF formats. It can analyse images visually if the selected model supports images. It cannot read older DOC files or scanned PDFs without a text layer - there is no OCR. The Assistant can read up to around 50 KB of text from a message attachment, up to 300,000 characters from a file selected as a knowledge source (in chunks of around 24,000 characters), and the knowledge base index stores up to 54,000 characters from a single file. Beyond these limits, the Assistant can only see the beginning of the document.