Marta runs a 14-person agency. On Monday a client asks why the scope in March's contract differs from the offer, and the answer is scattered between an email, a chat and a document nobody remembers the name of. A company's digital brain is meant to ensure that next time Marta finds the agreements where the team worked on them, and asks about them in a plain sentence.
A company's digital brain is a memory of the work, not another wiki
A company's digital brain is a way of organising work in which decisions and documents stay with the matters they concern, and the AI Assistant can answer questions from the available knowledge. It's not about an extra directory you have to fill in after hours. It's about keeping information on a change of project scope next to the conversation in which that change was agreed.
A wiki alone can store procedures well, but someone has to move everyday work agreements into it. A drive alone stores documents, but won't explain why the third version of the offer was chosen. So adopt three conditions: knowledge is created during work, it stays in one application, and AI reads it according to the permissions of the person asking.
A wiki empties out when conversations have to be copied into it
A company wiki loses its usefulness when recording a decision means a second action after the matter is closed. You can avoid this extra step by saving the agreement in the task or chat where the team is discussing it anyway.
Picture a Friday at Marta's. She asks the team to copy decisions about client ABC's project into the wiki; after three more Fridays she's filling in the entries herself, because the designers have already moved on to new tasks. So Marta goes back to the project conversation to reconstruct the decision that was missing from the entry.
Four layers are enough to start recording knowledge
Start with the matter, the current conversation, a lasting note and a document. Save each type of agreement where the team will look for it later.
| Layer | What goes into it | Example from Marta's agency |
|---|---|---|
| Matter | Description, deadline, people responsible and conversation about a specific outcome | The "Offer for client ABC" task with the agreed scope |
| Current agreements | A conversation still leading to a decision | The "Agreements with the client" chat with a question about the mock-up |
| Lasting knowledge | A procedure or summary you'll come back to for the next matter | A note from a meeting about project approval rules |
| Documents | Versions of offers, contracts and materials | A file with the scope of work attached to the thread |
In Task Manager, a task gathers the description, deadline, people and conversation; system messages record, among other things, a change of status or deadline. A Workspace groups tasks and chats, so you can treat it as a work area, for example "New website". You can read more about choosing where to put a decision in the article where to record team agreements.
A note is private until you share it. You can give a colleague viewing or editing rights; the Workspace label alone doesn't open access to them. If you're collecting agreements after client meetings, also see how to keep meeting notes. A separate Files module, with shared folders for files and notes, belongs to the Pro variant. Attachments in a task or chat stay with that thread.
How do you point to sources so the AI Assistant can answer?
To ask about a specific matter, open an AI Assistant thread and point to the materials it should use. In Task Manager you can combine several sources in a single question, without limiting yourself to one task or chat.
On Tuesday, Marta starts a conversation about client ABC's contract. As sources she picks the task with the offer, the chat with the designer, and a meeting note shared with her. Then she asks: "What did we agree on the scope of work?". Each conversation with the AI Assistant has its own history and set of sources; AI Assistant threads are private.
You can also point to your tasks, calendar or an entire Workspace. In the Pro plan, the source can also be a folder or a single file from the Files module, as well as your own mailbox from the Mail module - the latter only once the administrator has enabled mail as an AI knowledge source. Pointing to a task or chat covers files added to its thread. The usage bar shows how much of the selected knowledge will fit into a single question to the AI model. When there's too much material, the AI Assistant tells you what it didn't include. You'll find further examples in the description of the AI Assistant's knowledge sources.
In an open thread, you'll ask without the rest of the team being involved
If the answer lies in a single conversation, ask the AI Assistant directly in the task or chat. Only you see the question and answer, even though they relate to a thread where more people work.
On Thursday, Kuba comes back from holiday and opens the "Arrangements with the client" chat. Instead of asking Marta for a rundown, he picks the "AI Assistant" button and asks what was agreed; the private answer carries a "FOR YOU ONLY" badge. He can also use the "Organise" mode and the Summarise option to read a shorter version of the messages. This view only works during the session - it doesn't replace or rewrite the original conversation.
The Pro plan also searches without manually pointing to a document
The AI knowledge base in the Pro plan lets the AI Assistant find answers in the company's knowledge without pointing to a specific document each time. It's an index built in the background: tasks and chats go into it as summaries, while notes and files go in as fragments of the original content. The index is an organised collection of fragments that the AI Assistant searches through for answers.
Marta can ask: "What's our fuel allowance?", even though she doesn't remember the name of the note with the billing rules. With the default settings, the base refreshes once a day; the administrator can also trigger an update manually. Semantic search - that is, searching by meaning rather than just identical words - requires an OpenAI or Google Gemini provider and is enabled by default for them. Document fragments are then sent to the provider to generate vectors; these requests are billed separately. With Anthropic Claude and xAI Grok, indexing doesn't send material to the provider, although a regular question to the model still passes it the necessary context.
In the Standard plan, there's no AI knowledge base or semantic search toggle. The AI Assistant still answers based on the conversation and the sources pointed to in its window. The difference between these two ways of working is explained in more detail in the article about the AI knowledge base and knowledge sources.
Access to knowledge doesn't override permissions
The AI Assistant only uses material available to the person asking the question. Notes and Files keep their own permissions even in the Pro plan's knowledge base, including with respect to the administrator.
If you pin a private note to a task's folder, other participants won't see it just because they have access to the task. You need to share it with them explicitly. A file uploaded to the folder of the same thread, however, is visible to its participants. If you lose access to a source previously given to the AI Assistant, it will be skipped the next time you ask a question.
Mail (a module from the Pro plan) never goes into the background-indexed base; you'll point to your own mailbox manually only once the administrator has enabled this option. Application data stays on your server, but the question and the necessary source content are sent to an external AI provider using your company's own key. Before pointing to correspondence, check what gets sent to the AI provider with each question.
A scan won't answer, and a fresh decision may outpace the index
Task Manager won't read a PDF scan without a text layer, nor an old binary-format DOC file. For everyday work, you can use DOCX, XLSX and PDF documents containing text, and for a scan, save the needed arrangement in a note instead.
The second limit concerns how up to date the AI knowledge base is in the Pro plan. The automatic update runs once a day, so it's better to point the AI Assistant to a decision made an hour ago as a source, or ask about it directly in the thread. The administrator can also refresh the base manually. With long documents, keep the reading limit in mind: the index stores up to 54,000 characters from a single file, while a file pointed to as a source is read by the AI Assistant up to 300,000 characters, in parts.
A single place still makes sense even with AI switched off
You can build this working memory without turning on AI: Task Manager also works without it. Tasks, chats, notes and the calendar stay in the program, and the application data sits on your company's server. The Pro plan adds, among other things, Mail, Files and the AI knowledge base.
AI requires an account and your own key with the chosen provider: OpenAI, Google Gemini, Anthropic Claude or xAI Grok. You pay the provider directly, with no mark-up from Task Manager. The program licence is perpetual and paid once, based on the number of seats; it includes 12 months of updates, with later renewal optional. For a team of up to 20 people, the Standard plan costs 1,490 PLN, and Pro 3,990 PLN. If you're choosing between pointing to sources and a background-built base, also read the comparison of the Standard and Pro plans.
Tomorrow, start with one matter, not the whole archive
Tomorrow, pick one matter your team is handling: write its description into a task, leave the arrangements in the linked conversation, and attach the document it concerns. Then check whether someone else can reconstruct the decision without asking you about the project's history. If you want to keep application data on your own server and not pay a per-user subscription, see Demo.
Frequently asked questions
How do you start building a company's digital brain?
Start with a single matter: create a task, leave the agreements in a conversation and attach the document it concerns. Then ask another person to reconstruct the decision based on these materials.
Will the AI Assistant see everything in the company?
No, the AI Assistant uses materials available to the person asking. Notes and Files keep their own permissions within the Pro variant's AI knowledge base as well.
Does a company's digital brain require the Pro variant?
No, the AI Assistant and manual source selection also work in the Standard variant. The Pro variant is needed if you want to use the AI knowledge base built in the background and the Files module.
Will AI read old scans and PDFs?
The AI Assistant will read a PDF with a text layer, but it won't read a scanned PDF without text or an old DOC file. Task Manager has no OCR, meaning no function for recognising text in an image.
Does our data go to the AI provider?
Yes, the question and the required content from the indicated source go to the model provider using your company's key, although the application data stays on your server. Mail does not go into the knowledge base indexed in the background; as a manual source, it requires this capability to be enabled by an administrator.
Sources
The Task Manager Team - we write about teamwork, AI in business and keeping tasks organised, using examples from Task Manager, software installed on your own server. Task Manager has been developed by Blue Creation Group (Piotr Łącz) since 2013 and is used by organisations including Neonet, the Military University of Technology and Polpharma - read customer reviews. We check our articles against the user guide and the latest version of the software.
