AI is core to PartyGraph: the concierge that greets your guests, the moderation that protects them, and the song suggestions that help your DJ read the room. We want our use of it to be transparent, predictable and respectful of consent.
What the concierge can and cannot do
The concierge answers questions using a knowledge base that you, the host, provide — schedule, menu, venue, parking, supplier contacts, house rules. It is not a general chatbot. When it does not know an answer, it tells the guest and routes the question to your team.
The concierge does not invent schedule details, disclose private guest information, or make unmoderated introductions between guests.
Voice generation
The concierge's voice is generated using ElevenLabs text-to-speech models. Sample voice clips on this website are pre-generated and served as static files. No guest input is sent to ElevenLabs through any feature on this marketing site.
Models and providers
We use third-party models for natural language understanding, moderation and text-to-speech. The current providers are documented in our privacy policy. We may change providers in future; material changes will be communicated to hosts in advance.
No training on guest content
We do not allow our model providers to train general-purpose models on guest uploads, messages or feedback collected through PartyGraph events. Any future fine-tuning on aggregate event data will be described here in advance and subject to host opt-in.
Moderation
Automated moderation runs on photos, videos and text submitted via the concierge before they appear on a live screen or shared gallery. Hosts always have a final approval queue. We do not use facial recognition in the MVP product.
Limitations
AI gets things wrong sometimes. We design PartyGraph to fail gracefully: uncertain answers escalate, suspicious uploads are held for review, and the host can override every automated decision. The host remains responsible for the experience and content of their event.