Informace o AI asistentovi

Once activated, the built-in Yuteki assistant uses the Anthropic Claude API under the commercial services/API model. The operator must knowingly start the adapter after checking the account and settings. The default configuration does not allow real requests. Merely ticking the setting in the workspace does not bypass this block.

1. What the assistant is for

The assistant helps with projects and work organisation: it analyses the completeness of selected tasks, asks auxiliary questions, sends reminders about tasks without progress and proposes tasks based on permitted channels. It may automatically publish auxiliary comments; a proposal from the chat is turned into a task by an authorised user. Content originating from the assistant is marked in the interface.

The month planner prepares a proposed work plan for the team for the selected month. The distribution of tasks among people and days is calculated by an algorithm in the application; the language model is used only to estimate the time of tasks without an entered internal estimate. The proposal does not change the calendar or the tasks until an authorised person reviews it, accepts the estimates and applies the selected items.

AI is not intended for HR scoring, ranking employees or decisions on hiring, dismissal, promotion, remuneration, penalties or other significant personnel decisions. The results may be incorrect or incomplete. Verify them before using them in a significant decision; the name "assistant" is not a guarantee of correctness.

2. What data may be sent

• A selected task: title, description, type and status, acceptance criteria/checklist, information about assignees in the form of labels, deadlines and information on whether the task concerns a client.
• Selected messages from a permitted channel: content, author labels and time context. In the main agency-client channel this may include messages of client users. Not all channels or all messages are analysed automatically.
• Calling the assistant with @AI: the message with the call and the messages around it in the conversation in which the call was made, together with the replies in their threads. When the question concerns older matters, the assistant may review the history of that conversation from the last 6 months - it does not read older messages. This applies to every type of conversation, including private chats, client channels and project channels. The assistant may then also read selected fields of tasks within the scope of that conversation: title, description, status, type, deadline and priority.
• Estimation for the month planner: title, description, type and number of dependencies of open tasks without an internal estimate, in any open status, as well as the titles, types and hours worked of selected completed reference tasks. Client names are replaced with a marker, and known persons with labels. The following are not sent: people's availability and absences, rates, client quotes, prices or internal notes. Tasks excluded from the assistant, tasks of clients not covered by the assistant and tasks hidden by a project rule are estimated only locally.
• An instruction and a response schema defining the purpose of the analysis. Free-text content may contain personal data, names, addresses or business information entered by the user.

Without a call via @AI, private conversations, internal client channels and client project channels are not analysed. Internal task notes, time entries, settlement data and the content of attachments are not part of the payload of any of these paths. Reminders about tasks without progress do not require calling the model.

The application partially filters contact details and replaces known persons with labels in selected fields. This is not full anonymisation. Checklists, type and status names may be sent without this filtering; the filter also does not detect all data in a title, description or message. Do not enter unnecessary health details or content that may not be transferred to the provider.

3. What Yuteki stores

Yuteki stores the state and result of the operation, the generated proposals or comments, and usage metadata: the model, the number of tokens, the cost calculated according to the configured rates and the response identifier. Saved results, created tasks, notifications and usage settlements have a local life cycle separate from the storage of data by Anthropic.

4. Provider's rules

According to Anthropic's public information for commercial services/API, input and output data are not used by default to train models. There are exceptions related, among other things, to voluntarily submitting feedback, reporting errors or a conscious choice to allow the use of data. The settings of the specific account must be checked before activation. Anthropic's data usage rules.

Anthropic's standard information provides for deleting API inputs and outputs within 30 days of receipt or generation, with exceptions relating to separate arrangements, specific features, the law and the enforcement of safety and usage policies. This is not a guarantee of storage for exactly 30 days or a limit for every category of data. Current information on retention.

Data may be processed outside the EEA. Anthropic's public DPA describes processing and transfer mechanisms, including SCCs, for the agreements to which it applies. We do not declare EU-only processing or active Zero Data Retention. The description of the DPA itself does not confirm the terms of a specific account or settle Yuteki's transfer assessment. The provider also publishes a list of its subprocessors.

5. Control and limitations

New analyses require the appropriate permissions, current acceptance in the workspace, an active module and appropriate settings. The assistant can be switched off, or supported tasks, projects and channels can be excluded within the available scope. Permissions are also checked when a job is executed in the background. A participant of a client channel does not, by mere participation, receive their own switch that disables analysis for the whole channel.

Switching off AI stops new analyses in accordance with access control, but does not reverse completed transfers or delete earlier results. Access to saved proposals and the possibility of creating a task from them may remain after the module expires, subject to visibility and permissions. Acceptance in the interface does not replace establishing the right to use other people's data.

Availability also depends on limits, the network and the provider. A retry after an error may cause another request; we give no guarantee of single execution on the provider's side. An external assistant connected through MCP is a separate tool chosen by the user.

6. Contact

The service is run by Michał Piłat, an entrepreneur entered in CEIDG, NIP 9222839601, REGON 542359454. Place of business: Jatutów 90, 22-400 Jatutów, Polska. Address for service: Rabindranatha Tagore 5A/54, 93-503 Łódź, Polska.

Contact on matters of the service, privacy and security: biuro@yuteki.eu. Do not send a password or access tokens in your message.

Related information

Terms of service · Privacy policy · Cookie policy · AI notice