AI Use Disclosure

Reboko uses artificial-intelligence ("AI") models to help businesses reply to customer conversations. This page explains what AI does in Reboko, what data it can access, how you control it, and how we comply with App Store and Play Store AI-disclosure requirements (including Apple App Review Guideline 5.1.2(i) on explicit-user-permission for AI access to sensitive data).

1. What AI Does in Reboko

• Reply generation. When a customer sends a message on a connected channel (WhatsApp, Instagram, Messenger, Telegram, TikTok), Reboko can draft a reply and — with your configuration — either auto-send it or hand it to a human for review.
• Booking + order extraction. AI parses free-form messages into structured bookings, orders, and reminders.
• Summaries + escalations. AI summarises long threads and flags messages that need human attention.
• Knowledge-base lookup. AI queries your own knowledge entries (FAQs, menu, policies) to ground replies.

Reboko does not use AI to make binding legal, medical, or financial decisions on your behalf. AI-generated messages are suggestions; you remain responsible for what your business sends.

2. Where AI Runs

We use OpenAI (GPT family) as our AI provider via their API. Per OpenAI's API terms, data sent to the API is not used to train OpenAI's foundation models. Model choice may change over time; the disclosures on this page apply regardless of the specific model.

3. What Data AI Can Access

To generate useful replies, AI is given:

• The current customer message and recent conversation history.
• Business context you provide (business name, hours, reservation rules, services/menu, knowledge-base entries).
• Relevant operational data (open bookings, open orders) needed for the task.

AI does not receive:

• Your password, payment-card details, or device credentials.
• Other customers' conversations unrelated to the current thread.
• Content marked as internal-only (notes flagged as private in the dashboard).

4. Explicit Permission for Sensitive Data

Apple App Review Guideline 5.1.2(i) requires explicit user permission before AI accesses sensitive data. In Reboko:

• Connecting a channel (WhatsApp, Instagram, etc.) is an explicit opt-in step. Before connection, we explain that messages on that channel will be processed by AI on your behalf.
• Granting Contacts or Photos access (iOS / Android) is optional and tied to a specific user-initiated action. We do not passively scan address books or media libraries.
• Microphone + Camera are used only for user-initiated actions (voice note recording, QR-scan flow).

5. AI Labelling + Transparency

• AI-generated messages are clearly labelled in the Reboko dashboard so your team can distinguish AI replies from human replies.
• Where a connected channel (e.g. WhatsApp Business) requires disclosure of automated responses, Reboko provides configurable auto-disclosure templates.
• You can report an AI reply as harmful or incorrect directly from the thread; reports are reviewed by our team.

6. Your Controls

• Disable AI per channel. More → Channels → each channel has an "AI replies" toggle.
• Human-review mode. Configure the AI to draft but never auto-send; every reply requires a human tap to send.
• Escalation rules. Define keywords, topics, or customer lists that always route to a human.
• Report an AI reply. Inside any thread, long-press an AI message → Report. Our team reviews within two business days.

7. Limitations + Risks

AI output can be inaccurate, outdated, biased, or inappropriate ("hallucinations"). Do not rely on AI alone for safety-critical, legal, or medical communications. Review AI output before relying on it for anything where errors would cause real-world harm.

8. Changes + Contact

We will update this disclosure if our AI usage changes materially. Questions? Email [email protected].