What is Company Knowledge in Task Manager?
Company Knowledge in Task Manager is the material the AI Assistant uses to put together its answers. Instead of a generic answer, you get facts from your own work: a deadline from a task, a discount from a chat, a procedure from a note, or a rate from an XLSX file. You ask once, instead of searching in four places. You get the whole company's knowledge base, built in the background, in the Pro plan - together with Mail, Files, Automations and content encryption.
By default, the Assistant sees your tasks. You can add a whole workspace, a calendar (by default the last 30 and next 90 days), or a specific task, chat, note or files folder. If the administrator allows it, you can also add your own mailbox or a single email thread. You assign the selected sources to a conversation thread. They remain there for later.
In the Pro plan, a knowledge base indexed in the background runs alongside everything else. Tasks and chats go into it as summaries, while notes and files go in as fragments of the original. This means the database keeps the rates and clauses someone might ask about. The first fragment of every document is its title. So a question about the leave procedure might match the document's name, not just its content. Entry content is encrypted with AES-256-CBC. The knowledge base grows from material the team creates while working - with no need to enter knowledge separately.
How do you select materials?
Above the conversation is the Sources button. Browse the selection window just like the Files module: from the workspace through tasks and chats to folders, files and thread notes. Documents, meaning your private folders, and Mail are separate. Click a row to select a source. The › arrow takes you deeper, while “‹” goes back one level. The Search field at the top searches all sources at once.
Selecting a task or chat also includes its files: both those from messages and those added in the Files module. In the tree, you’ll find them under “Workspace / thread name”. The number of files is shown next to the thread name. You can only select your own folder from Documents as a source. It includes your files and notes from the entire sub-tree that are available to you. The Assistant reads long notes in sections and fetches more as needed.
You can assign up to 30 sources to a thread, with a content budget of 60 000 characters. Files attached to a task or chat have a separate limit of 8000 characters. The usage bar below the list shows how much space is used. If you select more material than the model can accept, the Assistant will tell you what it left out. General scopes are dropped first; explicitly selected tasks are kept. The programme checks permissions when saving sources and again when building a question for the model. If you lose access to a source, the Assistant will skip it the next time you ask a question.
What do you gain?
You don't have to dig through three conversations to find out what delivery date you promised a client last month. Ask one question and get an answer based on your own records. When you come back to it a week later, you won't have to gather everything again.
A new team member doesn't have to wait for someone to find time to explain the complaints procedure. They can get an answer from a company document without interrupting the rest of the team. You won't see a colleague's private note just because you work in the same workspace. You won't see it as an administrator either. Access to notes and files depends on whether they have been shared.
Use case: a quote for a client who got back in touch after six months
A client gets in touch after six months and asks whether the terms are still current. The original offer is in a task, the conversation about the discount is in a chat, and the PDF with the terms is in the thread folder.
You start a new conversation thread. At first, it can only see your tasks. You add a task and a chat as sources; their files are included without having to look for them separately. You ask about the previous terms and disputed points. The Assistant reads the task, messages and attachments, including PDFs with a text layer. Then you ask for a Word file with a summary. A download link appears in the conversation with the Assistant.
A scanned PDF with no text layer won't be read because the Assistant doesn't have OCR. It won't read an old DOC file either. If the terms are only in a photo of the signed contract, you'll need to transcribe them.
Frequently asked questions
Is the content of selected sources sent to an external AI provider?
Yes. The content of the selected source is sent to the AI provider chosen by the administrator in Settings → AI. Entries in the background-indexed knowledge base are encrypted using AES-256-CBC; vectors are not encrypted.
Can the Assistant access my email?
Only if the administrator enables the “Allow email to be used as an AI knowledge source” option in Settings → AI. It is disabled by default. Once enabled, everyone can only select their own mailboxes as sources. Nobody, including the administrator, can select someone else’s email. Email is never added to the background-indexed knowledge base.
Which files can the Assistant read?
The Assistant reads TXT, CSV, MD, JSON, XML, HTML, DOCX, XLSX/XLS and PDFs with a text layer. It analyses JPG, PNG, GIF and WebP images visually if the selected model supports images. It cannot read scanned PDFs without a text layer or older binary DOC files.
Does indexing Notes and files cost anything?
The AI model doesn't summarise them: excerpts of the original are added to the entry, and the program counts keywords itself. The knowledge base remembers up to 54,000 characters from a single file. With OpenAI and Gemini, when semantic search is enabled, the provider also calculates vectors for these excerpts - these are separately billed requests. Indexing starts together with the knowledge base update, so it requires configured AI.