Revinate Ivy (AI decision-intelligence layer)

par Revinate

AI Application

Faible confiance

Ivy is Revinate's decision-intelligence layer across its whole platform. Today it powers Chat (guest-facing AI concierge), automated scoring of every reservation call in real time (conversion opportunities and coaching moments), AI-written campaign content and segment-specific copy, sentiment analysis of reviews and surveys with suggested review responses, AI-driven identity resolution and data-quality checks in the CDP, and an 'agentic interface' exposing guest intelligence to customers' AI agents and other tools. Roadmap items include predictive insights, strategic campaign execution (creating segments, writing content, choosing channels and timing within guardrails), proactive occupancy monitoring and a continuous learning loop. It profiles individual guests at very large scale (the vendor cites 1.1 to 1.25 billion profiles across 12,500+ hotels) to decide who to target, with what offer, on which channel. Rated medium. Revinate claims SOC 2 Type II, GDPR, CCPA and PCI-DSS compliance, an isolated tenant, that all recommendations need customer approval, and that it never uses PII to train models without consent, but these were not verified (Trust Center and DPA unreadable). Key open points: Ivy 'learns' from behavioral data across all customers after identity is stripped ('synthetic data layer'), so de-identified guest data appears to be used for model learning by default, with no stated de-identification standard, consent mechanism or opt-out; the 'approval before execution' claim does not fit automatic outputs such as Chat replies, and the roadmap includes autonomous campaign execution; call scoring involves analyzing recorded calls and may be used to assess individual agents, with consent and retention not described; the 'agentic interface' exposes guest data to external AI agents with no stated access controls; model providers are not named; and no accuracy figures are published beyond an absolute claim that Ivy does not hallucinate. The rating would rise if the de-identification is not true anonymization, there is no opt-out, or agentic campaign execution goes live without approval gates.