Arbol’s artificial intelligence (AI) uses your patients’ data to serve them and to help your team. This page says what data it uses in each application, what Arbol’s policies say it doesn’t do with it, and what your team decides. If you haven’t yet, read AI at Arbol, in plain words first.
What the AI does with data
| Where | What data it uses | What for |
|---|---|---|
| Raíz’s agents | Their instructions, their knowledge and the conversation. On WhatsApp and web chat, also what’s already known about the person: their name, their captured data and the summaries of their earlier conversations. | To answer, and to carry out tasks with the tools of your organization’s integrations, such as checking slots or booking an appointment. |
| Raíz’s analysis | The conversation, when it closes. | To write its summary, outcome and mood, extract the Captured data and answer the agent’s evaluations. |
| Improve response, on messaging agents | The correction your team leaves on a reply. | To propose an improvement to the agent, which is tested against earlier conversations and applied only if someone approves it. See Improve an agent’s replies. |
| Arbi | What you can see in the connected applications, with your own permissions. | To answer what you ask. Before making a change, it asks you to approve it. |
| Ámbar | Information about your provider’s population. | To propose, for example, who joins a cohort. A person on the team decides. |
What the AI produces, such as a summary or a piece of captured data, is an aid to review quickly, not the record: when in doubt, read the transcript or listen to the recording.
What it doesn’t do, according to Arbol’s policies
These statements come from Arbol’s public pages:
- It doesn’t train models with your patients’ data. “We don’t use patient data to train third-party models or share it with advertisers. We never sell it.” (Privacy Policy, Colombia, in Spanish)
- Model providers don’t keep the data. “Model providers operate under zero retention and no training. What Arbol learns about your institution stays in your institution — and you switch that on yourself.” (Trust Center)
- It doesn’t use data for marketing. “We don’t use customer or patient data for marketing.” (Privacy Policy, Colombia, in Spanish)
- It holds no privileges of its own. “Arbol’s AI holds no privileges of its own: it gets no credentials and never reaches a database directly. Every action goes through tools that apply the same checks as any other access — organization, permissions, data ownership and consent.” (Trust Center)
- It isn’t a health professional. “Arbol isn’t a medical device and doesn’t provide health services: it doesn’t diagnose and doesn’t replace a professional’s assessment.” Nor is it an emergency line. (Terms of Service, Colombia, in Spanish)
You can also see this in the applications: on WhatsApp and web chat, Raíz’s agents have a rule never to ask for passwords, card numbers or verification codes.
What your team decides
- What an agent knows. Its instructions and knowledge are written by your team, and the agent uses them with everyone. Never put a specific patient’s data there. See Safe use and patients’ data.
- Notices to the patient. The Terms of Service say your institution is responsible for setting up the notices and obtaining the authorizations the law requires. In Raíz, what the agent says at the start is in its greeting and instructions: review them with the team at your institution that handles data protection. See Write an agent’s instructions.
- Whether calls are recorded. Each voice agent records audio only if Record calls is on, and it can hide numbers dialed on the keypad in the transcript. See Agent settings.
- When a case goes to a person. Clinical questions and cases that need judgment should be among those the agent hands to your team. The clinical decision always belongs to a professional.
- Which improvements apply. An improvement to an agent goes live only when someone with permission approves it.
For more on how Arbol governs its AI, read the Trust Center.