Elevate the guest experience and automate internal operations across all your systems.
Chat agents reply across email, WhatsApp, OTAs and social. Voice agents answer every call. Internal agents act inside your PMS and the tools your team already uses: dispatching work, updating calendars, closing the loop.
140+ languages, full observability on every agent step, SOC 2 Type II. 50M+ conversations and $3B+ in reservation value for 300+ brands including Marriott, Hilton, Nobu and Fairmont.
This is the 2nd launch from Conduit. View more
Conduit
Launched this week
Conduit automates guest conversations and the work behind it.
Conduit agents reply across email, WhatsApp, OTAs and social, voice calls. Agents dispatch subagents to act inside your PMS and tech stack: dispatching cleaners, updating calendars, closing the loop.
Agents learn from every interaction, and you can inspect every step. 140+ languages, SOC 2 Type II.
50M+ conversations and $3B+ in reservation value for 300+ brands including Marriott, Hilton, Nobu and Fairmont.
Free Options
Launch Team
Conduit
Adapted app
This is incredible. Congrats @punn_kam and the Conduit team!
@punn_kam I like that the goal isn't just faster replies but actually resolving requests. Have you seen that translate into better guest satisfaction scores?
the "calls nobody picks up at 11pm" line is the real value prop here, but that's also exactly when the riskier calls happen, a guest locked out with a medical issue, a smoke smell, someone who sounds actually panicked versus just annoyed about checkout time. does the voice agent have a hard trigger for escalating straight to a human on-call instead of trying to resolve it itself, or is that judgment call left to the model reading tone in the moment? for something touching door locks and payment systems already, I'd want the emergency path to be the most boring, hardcoded part of the whole system.
@omri_ben_shoham1 Yes, we have configurable escalation rules that allow customers to define topics that should ALWAYS be transferred to a human instead of being handled by AI. Common examples include lockouts, smoke or fire-related issues, and situations where a guest expresses frustration.
When one of these conditions is met, the AI immediately escalates the call. It also performs a warm transfer. During the transfer, it briefs the receiver on who the guest is and any relevant information the guest has already shared.
This creates a much better guest experience, especially in stressful situations because guests don't have to explain the issue again!
@omri_ben_shoham1 @emily_chau1 the warm transfer with context is the right shape for it, that answers the main worry. one thing I'd want to know for the 11pm case specifically: if the on-call human doesn't pick up on the first try, what happens next - does it ring a second person, keep the guest on hold with the AI in the meantime, or is there a point where it just has to attempt the resolution itself because nobody's answering? the escalation rule solves the "who" but the "what if nobody's there" case seems like the harder part of the same problem
Would love a guest-facing dashboard where travelers can view their full conversation history and any pending requests across channels in one place, would make handoffs feel way smoother when plans change mid-trip.
thanks Emily, the hardcoded escalation list plus warm transfer is exactly what I wanted to hear for a system already touching locks and payments. Gal's follow-up on the no-answer case is the harder half of this though - and the piece I'd add is what the guest actually hears while that's being sorted out. a panicked guest getting a vague "let me connect you" that just sits there is almost worse than no escalation at all, so I'd hope there's at least an honest "still trying to reach someone" update rather than silence or a loop.
Building for one industry properly beats building for everyone vaguely. Hospitality has its own rhythms and expectations a generic agent would fumble. Nice to see the focus.