Two Chains Just Published Their AI Concierge Numbers. That's Your Benchmark Now.

For three years the question was "does the AI concierge work?" Last week two of the biggest hotel groups in the world answered with numbers instead of adjectives. The question is closed. A new one opens: how do your numbers compare?

What they published

• Accor — 1 million+ guest conversations in the pilot since July 2025, now live across 5,800 hotels on its app, website, WhatsApp and iMessage, in 11 languages. 30–40% of those conversations are about booking — and the guests who book through the conversation convert better and spend more. A human is one tap away. And a second tool, "The Butler", does the same for staff.
• Wyndham — every call answered. An average property gets ~4,000 calls a year; industry-wide, about 25% are dropped. The AI picks up all of them. 85% of callers never ask for a human. Booking conversion on AI-handled calls is up 15%. And the calls that are handed to a person close at rates 16% higher.

Read that last pair together. The machine takes the volume; the human takes the calls worth a human. That's not AI replacing anyone. That's a wingman.

Why this matters more than the launch itself

The chains have chatbots since 2018. What's new is that they're now willing to put operational metrics on the record: answer rate, self-resolution rate, conversion, rate on escalated calls. Once numbers exist, they become the standard everyone gets measured against — including the 30-room hotel that never planned to compete with Accor.

The demand side confirms it. Accor cites Google's data: 60% of guests already use AI to search for trips, double a year ago. OpenAI's chairman told the industry last week that agents are "multilingual, with no hold time, and perfect access to information." The bar for "answered" has moved. A voicemail at 11pm is now a dropped booking.

The five numbers to steal

You don't need their platform. You need their scoreboard. Whatever answers your guests today — a person, a tool, both — these are the five figures the chains just told you to track:

• Answered rate. Of all calls and messages in a month, how many got a reply at all? Wyndham's baseline was 75%. What's yours after 8pm?
• Time to first reply. Seconds, minutes or hours. The first to answer still wins the undecided.
• Self-resolution rate. What share of questions closed without a human? 85% is the number on the table. Below 60% usually means your facts — breakfast times, parking, cots, policies — aren't written down anywhere a machine can read.
• Escalation quality. When a conversation is handed to a person, how fast, and with what context? Wyndham's +16% on escalated calls says the handoff is where money hides.
• Conversion of booking conversations. Accor's 30–40% booking share is a hint of how much of your inbound is revenue in disguise. Count it.

None of these requires a data team. A shared spreadsheet and thirty days will do. That's the Champions League an independent can play in: same scoreboard as a 5,800-hotel group, one property to fix.

The takeaway: for the next 30 days, log every call and message that reaches your hotel — answered or not, minutes to reply, closed by whom, booked or not. At the end you'll have your own version of the numbers Accor and Wyndham just published. Then decide what to hand the machine and what to keep human. The wingman handles the 85%; your team gets the 15% that were worth a human all along.