Canary AI Guest Messaging, Webchat and Agent Studio

par Canary Technologies

AI Application

Faible confiance

Canary's text-channel AI products: AI Guest Messaging and AI Webchat (a single AI model answering guests across messaging and chat channels and the hotel website, described on the AI Voice page as 'one AI model across channels'), and AI Agent Studio, which the homepage describes as 'Custom-Built AI Agents'. They sit alongside the Guest Messaging product, which is limited to 360 messages per room per year under the Terms. Rated medium, with low confidence: I read the navigation, homepage claims and the AI Voice page, not the product pages for these three, so what each does autonomously is not confirmed. The same findings apply as for AI Voice: strong assurance on paper (a public trust center listing SOC 2 Type 2 and PCI DSS, with gated DPA, sub-processor list and an EU AI Act transparency statement), but the public Terms and Privacy Policy say nothing about AI, the model providers are not named, de-identified customer data may be used to improve 'other Canary offerings' without an opt-out, and no accuracy or escalation metrics are published. AI Agent Studio, which lets hotels build their own agents, adds a configuration risk: the hotel decides what an agent may do and say, and Canary does not publish guardrails or approval controls for custom agents.