AI transparency statement

VoclyAI is an administrative communication system for licensed healthcare providers. It combines large language models with conversational logic, rules and escalation paths written for healthcare front desks. It works in Arabic, including Gulf dialects, and in English.

It runs on these channels, as each provider configures them: voice calls (inbound and outbound), WhatsApp, Instagram, TikTok, website chat, email, SMS. Inside the provider's shared inbox, Inbox Copilot assists staff; it does not write to patients on its own.

It is not a medical device. It is not used for diagnosis, treatment or clinical triage.

Every call and conversation starts with a disclosure that the patient is speaking or writing with an AI assistant, for example: "Hello, you have reached [Clinic]. I am an AI assistant…" The disclosure is system-level. A provider cannot switch it off.

Within the rules the provider approves, the system may:

• Answer administrative questions from the provider's approved knowledge base (hours, locations, services, doctors, insurance accepted, pre-visit instructions)
• Book, reschedule or cancel appointments in the provider's calendar, or capture the request for staff to confirm
• Route a conversation to the right team or branch
• Escalate to a person when asked, when unsure, or when a rule requires it
• Give the UAE emergency number 998 when a life-threatening situation is described

• Give medical advice, interpret symptoms or suggest treatment
• Pretend to be a person
• Answer from information that is not in the provider's knowledge base
• Override a patient's request to speak to a person

A patient can ask for a person at any time. The conversation is transferred with its context, or a callback is created for the team. Staff can take over any conversation in the shared inbox with one click and hand it back. Inbox Copilot drafts are reviewed and sent by staff. Every action is logged.

When a message contains red-flag symptoms, Inbox Copilot raises an Urgent Clinical Alert, blocks a generic administrative reply and starts the provider's escalation workflow, such as notifying the charge nurse and giving 998 guidance. These steps are hard-coded and cannot be disabled by staff.

Like any AI system, VoclyAI can make mistakes. The known limitations are:

• Rare incorrect or invented answers, reduced by answering only from the approved knowledge base
• Background noise, poor lines or unusual speech patterns that affect speech recognition
• Technical clinical terms or rare dialect words that are misheard
• Occasional misclassification of what the caller or sender wants

Fallbacks: when confidence is low the system asks again, offers a person or creates a callback. Flows are validated with the provider's team before go-live and reviewed afterwards.

The system uses the provider's approved knowledge base, the current conversation, and the connected calendar or clinic data the provider authorises. Data is hosted in the UAE. Patient data is not used to train AI models without the provider's explicit written authorisation. Details are in the privacy policy.

Under Federal Decree-Law No. 45 of 2021 on the protection of personal data, patients may have the right to know they are dealing with an AI, to ask for a person, and to access, correct, delete, restrict or object to the processing of their data.

Requests go through the healthcare provider, which controls the data. VoclyAI supports the provider in responding.

Questions about this statement can be sent to info@voclyai.com. We update this statement when the system changes in a way that affects what it does or how it is overseen, with a new version number and effective date.