Artificial intelligence in Quanty
Version 1.0 · 18.09.2026
This page explains where artificial intelligence runs in Quanty, who is responsible for what, and what the AI does not do. We publish it to meet the transparency duty in art. 50 of Regulation 2024/1689 (the AI Act) and so that our customers can show it to their own users.
1. Roles
We are the provider of the AI system: we build Quanty, choose the model providers, set the safeguards and place the system on the market under our own name.
Our customer is the deployer: it decides what to use the system for, how to configure the assistant and what data to put into it. Informing its own users and patients, and making sure its staff understand the limits of the tool, is its responsibility.
People who talk to the AI Secretary are told at the start of the call, before they give any information.
2. Who supplies the models
AI features in the application, that is chat, AI columns, and generating documents, presentations and dashboards, run on Anthropic Claude models served by Amazon Bedrock in Amazon Web Services. We also use models there for vector embeddings, for reranking search results and for transcription.
Web search started by a user is run by Exa.
AI Secretary calls are handled by Eleven Labs: speech recognition, running the conversation, speech synthesis and the summary. The language model that holds the conversation comes from Google (Gemini, and Claude served in Google Cloud) or from OpenAI (GPT models), depending on the configuration and on the backup model.
The full list of providers, processing locations and transfer safeguards is at /subprocessors.
3. What the AI does
It reads the content you point it at and answers on that basis.
It fills columns in tables following your instruction, for example classifying, summarising or pulling a value out of a document.
It drafts documents, presentations and dashboards.
In the AI Secretary it holds a phone conversation, gives organisational information and writes the request into a table.
4. What the AI does not do
It does not decide for you. The output is material to check, not a ruling.
It gives no medical, legal, tax or investment advice.
It does not assess symptoms, does not decide how urgent a matter is, does not determine eligibility and does not build priority lists.
It does not recognise a person by voice. We create no voiceprint or other biometric profile.
It does not profile people and takes no decisions producing legal effects for them.
It does not learn from your data. We do not use customer content to train models.
5. A person checks the output
Language models can be imprecise and can state something that sounds credible and is untrue. That is true of every provider.
So any output that would justify an action has to be checked by a person. In the AI Secretary a request is always handled by the clinic’s staff, and an appointment is confirmed by a person.
We mark AI-generated content in the interface. Extending that marking to exports is planned.
6. AI Secretary: what the caller hears
At the start of every call the assistant says that it is a program based on artificial intelligence and on whose behalf it acts. That announcement cannot be switched off or shortened in the configuration.
The assistant says that the call is stored and processed by the organization and its technology providers, including outside the European Union, and points to where the full privacy information can be found.
Asked whether the caller is speaking with a human, the assistant answers truthfully.
Where a situation may indicate danger to life, the assistant asks the caller to hang up and dial 112. It does not assess the caller’s condition and does not connect to the emergency number.
A caller can ask to speak with a member of staff.
7. The AI Secretary is not a medical device
The AI Secretary is an administrative tool for taking phone requests on behalf of an organization: it gives organisational information such as opening hours, address, the range of services and prices, collects contact details and a description of the matter, and writes them down for the organization’s staff to handle.
The AI Secretary is not a medical device and is not used for diagnosis, symptom assessment, urgency assessment, eligibility decisions or medical advice.
We do not build, and will not build, a feature that routes a caller to a specialist based on symptoms without a documented regulatory review first.
8. How to report a problem
If AI in Quanty answered incorrectly, went beyond its role or behaved in a way that gives you doubts, write to dawid@pluscode.io. Give the date, the organization and, if you can, the part of the conversation or the request identifier.
Security reports are handled as described at /security.
Reports about the behaviour of the phone assistant are treated as a priority and may lead to that assistant being paused until the matter is clear.
9. What we do not claim
We do not claim to be "AI Act compliant". No certification exists for systems covered by art. 50. We describe what we do.
We do not claim that AI Secretary processing takes place only in the European Union.
We do not call ourselves a partner of any model provider.
10. Changes
A change of model provider or a material change in the assistant’s behaviour is recorded here and, where it concerns sub-processing, at /subprocessors.
The Polish version is the binding one.
