On Monday, Marta returns after three days out of the office. She opens the “Client arrangements” chat: 42 messages, and somewhere among them a changed deadline, a decision on a dark theme, and an unanswered question. She needs the outcomes, not the whole transcript. An AI conversation summary can help her find the decisions, deadlines and people responsible, if they were mentioned in the thread.
Why does the question “where did we leave off?” require so much searching?
Because decisions get mixed up with questions, short confirmations and attachments. When the outcomes are still scattered across several channels, reading a single thread still won't give you the full answer.
Marta finds confirmation of the dark theme in the chat, but she remembers the deadline change from a phone call. No chat summary can bring that back. Before you ask AI for help, gather the outcomes relating to the case in one place. After the call, Marta adds the decision about the deadline change to the thread. The text on why teams lose track of agreements shows where this problem comes from.
Not every case needs a separate chat. Sometimes it's better to hold the conversation within a task with a due date and an owner, and sometimes a meeting note is more useful. If you're wondering where to record your team's agreements, choose the place the person doing the work will come back to. AI can help you read the record; it won't replace the record itself.
How do you get an AI conversation summary you can actually act on?
Ask for three sections: “Decisions”, “Deadlines” and “Who's doing what”. This layout separates what's already settled from work that's still waiting for an owner.
Instead of just writing “summarise”, Marta asks: “Summarise this thread. List the decisions, deadlines, and who's responsible for what. Point out any unanswered questions.” In the example case, she'd see the decision on the dark theme, the mock-up deadline and the name of the person preparing it. If the conversation doesn't contain confirmation of the new deadline, she treats that gap as a cue to follow up, not as the client's agreement.
Start reading from the most recent arrangements. Friday's message may have changed Tuesday's decision; comparing the two without checking the order can easily mislead you. Then take one action: add the missing decision to the conversation, or turn the open matter into a task.
In Task Manager, you ask the AI Assistant right where the conversation is happening
In Task Manager, you can ask for a summary without copying the thread into a separate chat. The “AI Assistant” button next to the reply field in a task or chat switches on a private question mode.
Marta clicks it in the “Client arrangements” chat. The editor turns purple, and the bar above it states that participants won't see the question. She types her request for decisions, deadlines and owners. The answer appears in the same thread with a “ONLY FOR YOU” badge; neither other participants nor the administrator see the questions or the answers. Clicking “AI Assistant” again switches back to writing regular messages.
The AI Assistant has access to the thread's content and its attachments, provided you have access to them too. It can read, among others, DOCX, XLSX and PDF files with a text layer; it will analyse an image visually if the selected model supports it. This is useful when the decision was made in an attached document rather than in a message. The rules governing this are described on the AI Assistant in Task Manager page.
The question being private doesn't mean the query stays on the company's server. When you use AI, the content of the submitted thread is sent to the provider chosen by the company. This should be distinguished from manually pasting a conversation into a separate tool: here you're asking within the case itself, but the model still processes the context sent to it.
Four modes help you read a thread, not just summarise it
If you need to go through successive messages in a shorter form, choose “Organise” rather than a private question about a single answer. The button offers four modes for reading the same messages: Original content, Polish, Summarise and Expert.
| Mode | What you'll see |
|---|---|
| Original content | Messages without AI processing. |
| Polish | The same content presented more briefly and neatly, in the first person. |
| Summarise | Each processed message captured in 1-2 sentences. |
| Expert | A professional version with an “Expert note” line: advice, a risk, or a question to clarify. |
The AI starts with the 8 most recent messages. If you need older ones, use the “Organise older messages (N)” button under the mode banner. Within the same session, you can switch between already prepared views without waiting again. Under the processed messages you'll see the note “AI can make mistakes”.
On Tuesday, Marta chooses “Summarise” to go through the client's replies. Later she switches to “Expert” and checks which questions about the mock-up need clarifying. She treats these notes as suggestions, not as new agreements with the client.
This view isn't saved: once the session ends, you need to set it up again. On the other hand, “Organise” doesn't overwrite the originals. If someone questions a deadline, you go back to the source message instead of guessing whether its wording was changed by AI.
If the findings need to stay, turn the answer into a file or a task
Save a summary you want to come back to as a file, or turn the work it implies into a task. The “Organise” view itself is for reading during the session, not for storing a new version of the conversation.
In a private question mode, Marta types: “Prepare a Word document summarising this thread.” AI puts together the content, and Task Manager builds the file and shows a “Download …” button; this way you can also get a PDF or HTML. A single message, including an AI reply, can be exported from the ⋮ menu to PDF, Word or TXT. If you export a message viewed in “Organise” mode, the file will contain the version processed by AI, with an appropriate note. You can find out more about how to keep agreements in a PDF in a separate example.
A file preserves the state of the matter, but it doesn't assign work. So on Wednesday Marta asks: “Create a task: mock-up corrections, deadline Friday, assign to Marek.” The AI Assistant opens the “Add Task” window with the fields filled in. Marta checks the content and the deadline; the task is only created once she clicks “Save.”
You don't need to open the thread, but you do need to point to the right source
In the AI Assistant window you can point to a specific task, chat or note as the source and ask for a summary without opening the conversation. If you point to a task or chat, the AI Assistant also picks up the files attached to it.
On Friday, Marta points to the “Recruitment” task as the source and asks: “Summarise the Recruitment task.” Still check whether you pointed to the right thread and whether the answer covers the decision you're looking for. The Assistant only sees content you have access to; it won't find a phone agreement that nobody wrote down.
The same “AI Assistant” and “Organise” buttons are also available in an email thread. The Mail module is only available in the Pro plan, so don't plan this way of working with the shared inbox in the Standard plan.
AI needs configuration and won't read every attachment
The AI buttons appear once the administrator has configured the service and you have AI permission in the given context. Reading attachments has its limits: a PDF scan without a text layer and an old DOC file won't be read. Task Manager also works without AI; tasks and chats don't require a model connection.
A company chooses OpenAI, Google Gemini, Anthropic Claude or xAI Grok and uses its own API key, meaning the access credentials for the provider's service. For using the model, it pays the provider directly, with no mark-up from Task Manager. The application's own data sits on the client's server, but the request with the context sent to the model leaves that server. If you're discussing confidential matters, read carefully what reaches the AI provider.
There are also limits to reading. Task Manager won't recognise text from a PDF scan without a text layer, because it has no OCR, meaning the function of reading text from an image; it also won't read an old binary DOC file. If the agreements are in the body of the conversation, the attachment reading limitation won't stop you from summarising them. When the model doesn't respond, the program shows the reason, for example an invalid API key, a request limit, or an unavailable model.
Tomorrow, pick one long thread. Ask for “Decisions”, “Deadlines” and “Who's responsible for what”, check the answer against the originals, and turn one ownerless matter into a task. In Task Manager, the conversation, the private question to AI and the task remain part of the same piece of work. The program is available under a perpetual licence with a one-off payment per number of seats, with no per-user subscription. If this setup suits your team, check out the Demo.
Frequently asked questions
Will others see my AI conversation summary?
A private question asked via “AI Assistant” in a thread, along with the answer marked “ONLY FOR YOU”, is visible only to its author, not to other participants or the administrator.
Will “Organise” change the original messages?
No. “Organise” modes create a view for the duration of the session; they don't save the processed version or overwrite the originals.
How do I keep a thread summary for later?
Ask the AI Assistant in the thread for a Word, PDF or HTML summary file. You can also export a single reply from the ⋮ menu to PDF, Word or TXT.
Will AI also summarise an email conversation?
Yes. The “AI Assistant” and “Organise” buttons are also available in Mail, which is only available in the Pro plan.
What should I do if a thread has 60 messages?
In “Organise” mode, start with the 8 most recent messages, then pull in older ones with the “Organise older messages (N)” button. If you mainly need decisions and deadlines, ask the AI Assistant a private question about the thread and check the answer against the originals.
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.
