AI Transparency Policy
§ 1. Purpose of the Document
This Policy explains how artificial intelligence technology is used in the Boostra Application and the rules for handling data transferred to the models. It supplements the Privacy Policy and the Terms and Conditions.
§ 2. Use of Artificial Intelligence
The Application uses artificial intelligence models to analyse business and sales data, analyse marketing activities, identify problems, generate recommendations, create reports, generate content, monitor indicators, and operate the chat. The system performs a supporting function and does not replace human judgement. Within the meaning of Regulation (EU) 2024/1689 (the AI Act), Boostra is the provider of the AI system, because it makes the Application available under its own name, while the Client using the Application is the deployer of that system. The Application does not belong to any of the categories listed in Annex III to the AI Act, nor does it constitute a safety component of a product within the meaning of Article 6(1) of the AI Act, and accordingly it is not a high-risk AI system.
§ 3. Content Generated by AI
Some of the content available in the Application is generated by artificial intelligence models, in particular recommendations, reports, summaries, analyses, forecasts, action plans, optimisation suggestions, and product descriptions and marketing materials. This content is marked in the Application's interface. The User should verify it before use.
Content generated by the Application is made available as drafts. It is saved, exported or published to a connected channel only after the User has previewed it and given express approval. Images generated or modified by the Application are marked in a machine-readable format, in accordance with Article 50(2) of Regulation (EU) 2024/1689. For text content, we apply a marking in the Application's interface; we do not apply a persistent machine-readable marking to text, because the solutions currently available do not allow such a marking to be preserved for this type of content. The cited provision requires that, when choosing solutions, account be taken of the specifics and limitations of each type of content, the cost of implementation, and generally acknowledged technical know-how. The scope of the markings applied is reviewed as available solutions develop.
§ 4. Chat and Notice of Interaction with the AI System
A User using the chat is informed that they are interacting with an artificial intelligence system and that the responses are generated automatically on the basis of data from connected integrations.
§ 5. Human Oversight and the Limits of Automation
The Application does not independently make legal, financial, tax, accounting or HR decisions. It does not make decisions concerning natural persons based solely on automated processing that would produce legal effects concerning them or similarly significantly affect them.
Without the User's approval, the Application performs only operations that do not change the state of the User's external accounts: data synchronisation, reading statistics, calculating indicators, generating recommendations and alerts, and preparing content drafts. Every write operation in an external system requires the User's express approval after the object, the proposed change and its effect have been presented.
The Application maintains a log of the actions performed, covering the User, the time, the platform, the operation identifier and the result returned by the external interface, and, where technically available, the object and its state before and after the change. A completed action cannot be undone. A change made in an external system can only be superseded by a further action, for example pausing or resuming a campaign again, or changing a product description again. The log serves to reconstruct the course of actions, not to automatically reverse them.
§ 6. Limitations of AI Systems
The User acknowledges that artificial intelligence models may make mistakes, generate incomplete responses, rely on incomplete data, produce outdated recommendations, and may not take all business circumstances into account. Boostra does not guarantee the correctness, completeness or usefulness of any generated recommendation.
§ 7. Data and Model Training
1. Boostra does not use Users' data, or data retrieved from their stores and accounts, to train, fine-tune or validate artificial intelligence models - whether its own models or those of third-party providers. This prohibition also covers pseudonymised data.
2. Boostra does not reserve the right to use data in this way, and does not make it conditional on the Client's consent.
3. In its cooperation with model providers, Boostra uses solutions that exclude the use of the data transferred for training those providers' models, and that limit the retention period for queries.
4. Boostra may use synthetic data and data that has been completely and irreversibly anonymised, including in aggregate summaries and industry comparisons made available to other Clients. This is conditional on it not being possible to identify a person or a specific store, on an adequate minimum group size, and on it not being possible to reconstruct the source data. This does not apply to the training of models.
5. Access to the content of conversations and to a specific Client's data may be obtained only by a limited group of authorised technical and support personnel, and only where this is necessary to handle a request or an incident. Access is time-limited and logged.
§ 8. AI Technology Providers
To provide the Service, Boostra uses models made available by third parties. The current list of providers, together with an indication of their function, processing region and the basis for transferring data outside the European Economic Area (EEA), is set out in Annex No. 3 to the Data Processing Agreement. We inform Clients of an intended change of provider at least 30 days in advance, via a notice in the Application or by e-mail.
§ 9. Location of Processing
The Application's production environment, database and backups are maintained with an infrastructure provider within the European Economic Area. System administration is carried out from the territory of Poland. Some model providers may process data outside the European Economic Area, in particular in the United States. The basis for the transfer for each provider is indicated in the list referred to in § 8.
§ 10. Scope of Data Transferred to the Models
In queries sent to the models, we transfer the data necessary to answer the User's question: aggregated sales and marketing results, product data, and business context. The Application does not transfer a buyer's first name, surname, e-mail address, telephone number or address to the models. If the Application presents the User with information concerning a specific buyer, the linking of the result to identifying data takes place on the Application's side, within Boostra's internal tools, without transferring those identifiers to the model provider. The Application does not, however, block content entered independently by the User. If the User enters identifying data of a buyer into a query, in particular a telephone number, e-mail address or first name and surname, that data will be transferred to the model provider together with the query. The User is responsible for the scope of the data entered into queries; please do not include buyers' identifying data in them.
§ 11. Prohibited Uses
The User may not use the Application for actions that violate the law, in particular for discrimination, infringing the rights of third parties, making automated credit decisions, making HR decisions without human involvement, or actions contrary to the regulation on artificial intelligence.
§ 12. Contact
For matters relating to the use of artificial intelligence: kontakt@boostra.pl. BOOSTRA Sp. z o.o., ul. Sudecka 153, 53-128 Wrocław.