Monday, 8:40. Ola returns from leave to her seven-person accountancy firm, Bilans Plus, and finds 46 messages waiting for her in one thread. For Ola, AI in business only makes sense if it helps turn that number into decisions and deadlines. According to Statistics Poland, 8,7% of companies employing at least 10 people used AI in 2025, compared with 6,1% of small companies employing 10 to 49 people. Instead of another presentation about artificial intelligence, you will see seven specific situations, each with a prompt, result and cost.
AI in business makes sense when it works with the right context
AI saves businesses time when you do not have to manually copy emails, tasks and decisions into it. In Task Manager, you select the sources and the AI Assistant responds based on content you have access to.
A standard chat open in another tab does not know who the client Nord is, what deadline was agreed last week or why Marek is waiting to work on the mock-up. So first you copy the conversation into the model, then transfer the response back. A short question may only take a minute, but with 20 such operations, you create another chore rather than saving time.
In Task Manager, you can select your Tasks, Calendar, an entire Workspace, a specific task, chat, note, folder, individual file or your own mailbox as a source. A bar shows how much information was included in the request, and the AI Assistant tells you what it did not include.
You choose the provider: OpenAI, Google Gemini, Anthropic Claude or xAI Grok. You enter your own API key, and the list of models is retrieved from your account. You can learn more about this mechanism on the AI Assistant in Task Manager page.
Seven Use Cases for AI and Automations at work
The AI Assistant summarises conversations, creates checklists, prepares tables with files, translates messages, sets reminders and proposes AI Agent plans; a separate scenario shows an Automation Recipe that creates a task when an email arrives. These are seven separate situations from Bilans Plus, Studio Kreska and Termo-Serwis, not a plan for a single week.
The AI Assistant prepares content and forms, but you save the task, confirm the email or enable the Automation Recipe.
1. You return from leave to find 46 messages waiting in a thread
On Monday, Ola opens the “Client arrangements” conversation. Instead of reading everything in order, she selects Summarise in Organise, which reduces each message to 1-2 sentences and processes the conversation in batches of 8 messages.
She can also ask privately in the thread: “Summarise this conversation. List the decisions, deadlines and people responsible.” Only she will see the response, marked “ONLY FOR YOU”.
The Organise preview disappears at the end of the current session, while the original messages remain unchanged. The application does not save a new version of the conversation.
This is how AI conversation summaries work: first you get an overview, then you open only the messages that require a decision.
2. A two-page brief becomes five points
On Tuesday, Michał from the twelve-person Studio Kreska opens the “Website project” task. The description is two pages long, so he enters a private prompt: “Turn the task description into a checklist that I can tick off.”
The AI Assistant returns five points, but its response is still private. Michał therefore adds them to the task description as a to-do list. From then on, the team can tick off items without opening the editor, a bar shows their progress, and each completed item shows who ticked it off.
When Ania finishes the mock-ups and ticks off her item, everyone else sees the same status. The difference is simple: the checklist in the AI Assistant's response is for Michał, while the to-do list in the task description is for the whole team.
3. A complaint received by email becomes a task
At 16:47 on Wednesday, a complaint arrives at the nine-person Termo-Serwis. It is not handled independently by the AI Assistant, but by a previously saved and enabled Automation Recipe.
The Recipe responds to a new email and checks a condition described in plain language. It then creates a task in the Sales Workspace for Kasia Wójcik, with high priority and a due date one day later.
Automations follow a “WHEN - IF - DO” structure and run within the application. A rule can take the sender, subject, mailbox and presence of an attachment into account, and one Recipe can handle up to 10 actions.
Automation safeguards
Before running it, you click “Test” to see the expected result without performing any actions. The history covers 30 days and shows reasons for skipped actions and any errors. After three consecutive errors, the Recipe stops and sends an email. An email rule responds only to its author's mailboxes.
Mail and Automations are included in the Pro plan. The full workflow is shown in the example of an email that creates a task.
4. The report goes into Excel without manual data entry
Michał asks the AI Assistant for a report from a selected Workspace. After checking the table, he adds: “Generate an Excel file from this”.
AI prepares the content, and the application generates the file. Available formats are PDF, DOCX, XLSX, CSV and HTML. The private link works for 24 hours and can only be opened by a signed-in user who has access to the source task or chat.
For a large project, Michał selects a specific Workspace or individual tasks as the source instead of asking about everything at once. The source bar shows how much content fitted into the query, so he does not have to guess whether the report covered all the materials.
5. A German tenant writes to Polish technicians
On Friday, Termo-Serwis receives a message from the hall manager. Kasia asks in the thread: “Translate the latest message into Polish”, and once she has prepared her reply, she asks for it to be translated the other way.
The AI Assistant can see the selected conversation context, so it can retain equipment names and previous arrangements. The translation has a “JUST FOR YOU” label. The client only receives the text after Kasia has checked it and copied it into a standard message.
6. One sentence keeps track of an invoice deadline
Ola types: “Remind me tomorrow at 9 about the invoice for Nord”. The other type is a reminder linked to a task's due date, such as one day before it - if the due date changes, the reminder moves with it. Up to 50 reminders can be active at the same time.
Once a day, the AI Assistant sends a summary showing the number of tasks due today and overdue. Separately, the Notification Centre reminds you 15 minutes before an event, when it starts and one hour before a task is due, and displays a morning briefing with the day's agenda.
In-app reminders always work. Emails sent in the background while the application is closed require CRON, which is a scheduled task run on the server. The program warns you about this when you set a reminder.
7. The AI Agent creates a plan instead of a single answer
Ola is waiting for Kasia's reply about the proposal for Nord. She enters a complex instruction: “If Kasia hasn't replied by tomorrow, remind her at 8:30 and let me know at 9:00”.
The AI Agent is an AI Assistant mode designed for multi-step tasks. It proposes a plan that appears in the thread and operates strictly within the permissions of the person who started it. Delayed steps require CRON on the server.
This is more than a text response, but you still see the plan before it takes further action. This mode is described in detail on the AI Agent page.
What requires your confirmation and what does not change the program
Text responses from the AI Assistant do not change anything in the program. You save the task, confirm that the email should be sent and enable the Automation Recipe.
| What AI prepares | What happens next |
|---|---|
| A summary, table, translation or checklist in the thread | The response remains private content and does not change anything |
| Task from the Assistant | The AI Assistant opens a form with the fields completed, and you save the task |
| Email from the Assistant | The message is only sent after you confirm it |
| Automation rule | You save and enable it, while “Test” does not perform any actions |
| Reminder or AI Agent plan | It is created from your instruction and is visible in the thread |
Once enabled, the Automation performs the saved actions, such as creating a task when an email arrives. It uses the author's permissions, has an hourly run limit and includes protection against loops.
How much AI costs your business: one-off licence, tokens charged separately
The cost of AI in Task Manager consists of the program licence and optional model usage charges. You pay OpenAI, Google, Anthropic or xAI directly for tokens using your own API key, with no Task Manager mark-up.
The licence is perpetual, and its price depends on the number of seats. You do not pay a subscription per user. The price includes 12 months of updates. You can then renew access to new versions or keep your current version without any further licence fees.
| Plan | Standard | Pro |
|---|---|---|
| Starter, up to 8 people | 690 zł | 1 990 zł |
| Classic, up to 20 people | 1 490 zł | 3 990 zł |
| Max, up to 35 people | 2 290 zł | 5 990 zł |
The Pro plan adds Mail, Files, Automations, AES-256 content encryption and the AI knowledge base, which lets the Assistant answer using material from the whole company. The AI Assistant itself works in both plans. For teams above 35 people, Enterprise is available with individual pricing.
Providers charge separately for input and output tokens. A token is a piece of text processed by the model. Anthropic estimates that one token corresponds to approximately four characters in English, while Polish text usually requires more tokens.
Assuming 8000 input tokens and 600 output tokens, a single summary may cost approximately 0,003 to 0,02 USD, depending on the model. A team of 10 people, each asking 10 such questions over 22 working days, will generate 2200 queries. The estimated bill will range from a few dollars to approximately 44 USD per month. You set the spending limit in the provider's dashboard.
The program also works without AI, so model charges are optional. Billing is explained in more detail on the AI costs with no mark-up page.
Where your data goes when you request a summary
Application data, such as tasks, messages, files and mail, is stored on your server. However, when you use the AI Assistant, the content selected as context for the query is sent to the chosen model provider using your API key.
The administrator enables AI separately for tasks, chats and the AI Assistant window, while the user selects the sources needed for the response. Mail is disabled as a knowledge source by default. Once it is enabled, the AI Assistant and Automations can only access the user's own mailboxes.
The AI Assistant only uses content that you can access. The global administrator manages the application and can see tasks and chats, but mail, notes and conversations with the AI Assistant remain private, even from them. API keys and SMTP passwords, which are the login details for the mail server, are encrypted in the database.
The server connects to the services that have been enabled, e.g. an AI provider, a mail server, Google Calendar or n8n webhooks. Without a separate decision, it only connects to the licence panel, to which it sends licence data and the program version, without any content from the database. The browser loads libraries, fonts, icons and avatars from the application server, not from external services. This means the browser itself does not fetch these elements from third-party servers.
The AI Assistant reads TXT, CSV, MD, JSON, XML, HTML, DOCX, XLSX and PDF files. The program does not have OCR, which is the ability to recognise text in scans, so it cannot read a scanned PDF or a document in the old DOC format. It can analyse images visually if the selected model supports this feature.
Before connecting a provider, check its data processing terms. You can configure the provider, key, model, response length limit, conversation history and attachment summaries under Settings → AI.
Start with one thread tomorrow
To begin, choose a thread that took you more than five minutes to read today. Tomorrow morning, ask the AI Assistant in that thread: “What was agreed, who is responsible for what and what deadlines were mentioned?”.
Do not configure seven scenarios straight away. Test one result using your own data in the Demo.
Frequently asked questions
Can AI in my business do anything without me clicking?
The AI Assistant does not save anything for you - a task is created in a form that you approve, and an email is sent only after you confirm it. The only things that work without you clicking are a previously saved and enabled Automation, and a reminder or AI Agent plan that you have commissioned yourself.
Which AI model should I choose to start with?
Start with a cheaper model available on your account and test it on summaries, checklists and translations. For long documents, you can compare the result with a more advanced model by changing one field in Settings → AI.
Is data from my server sent to the AI provider?
Yes, the content of the request and the context selected as a source are sent to the provider. Your application data remains on your server, but the request is processed by the selected provider using your API key.
Do I have to pay for AI to use Task Manager?
No, Task Manager works without AI. An account with a model provider is optional, and once connected, you pay the provider directly with no intermediary markup.
Will an AI Agent plan work after I close the browser?
No, unless CRON is running on the server. Delayed AI Agent steps and emails processed in the background require CRON, a recurring server task that triggers actions without an open browser.
Sources
- Statistics Poland, “Information society in Poland in 2025”
- OpenAI, “Pricing”
- Anthropic, “Pricing”
- Google, “Gemini Developer API pricing”
- Task Manager, pricing
- Task Manager, “AI Assistant”
- Task Manager, “AI Assistant”, “AI settings” and “Automations” guides - https://taskmanager.pl/materialy/asystent-ai, https://taskmanager.pl/materialy/ustawienia-ai, https://taskmanager.pl/materialy/automatyzacje (accessed: 2026-09-22)
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.
