How to use AI to make more from late hotel check-outs

Offering a late check-out should be simple. Guests want it, and they’re often happy to pay forty or fifty dollars for it. So why do so many properties have to say no? 

It’s not because the room isn’t available. It’s because they don’t have the system in place to connect the operational pieces together. 

That's the story Cole Rubin, Co-founder and CEO of Conduit, told Matt Welle on the latest episode of Matt Talks Hospitality. Before building an AI agent platform for hospitality, Cole ran a $19 million residential portfolio of around 200 properties across Airbnb and Booking.com.  

Cole knows the midnight messages and the missed calls firsthand, because he used to be the one answering them. Here’s how he fixed the problem. 

The forty-dollar problem 

Cole says many hotels lose money orchestrating and facilitating these early check-outs, to the point where they choose not to offer them. A request that should be a quick yes turns into four separate steps of back-and-forth between front desk, housekeeping and the guest. 

Conduit's approach is to let an AI agent handle that whole chain: checking the reservation calendar, confirming with housekeeping, taking payment, updating the guest, all without a human in the loop at all. It’s indicative of the wider shift Cole and Matt spend the episode discussing: guest communication is no longer the endpoint of an AI agent's job. It's the trigger for everything that happens next. 

From unified inbox to back-office agent 

AI can only work effectively if it has the full context of what’s needed. That’s why Cole and Conduit team started with the infrastructure around guest communication. 

Their first year was spent unifying every channel a guest might use: texts, Airbnb messages, booking site inboxes and PMS data. An agent that can only see half a guest's conversation can't act on it. Only once this was all in place did the team turn to the AI itself. 

The result? "We can instantly respond to that message – most times – better than a human would able be able to." When the agent can't handle something, it hands off to a person through the same inbox. And increasingly, it doesn't stop at the reply. It sends the payment link, updates the checkout time, and notifies the cleaning team. In other words, it handles the tasks that accompany the message, not just the message itself. 

Teaching the machine, not fixing the message 

AI will sometimes make mistakes – just as humans do. The trick is ensuring the mistakes don’t happen again. 

When the AI gets something wrong, don't just patch the individual message and move on. Go back to the input, the knowledge base, the instruction, the guardrail, and fix the thing that caused the mistake so it doesn't happen again. Conduit has given that role a name: conversation engineer. Someone whose job is to sit with the conversations the AI couldn't handle and teach it what it was missing. 

It's the same discipline that good managers use with new hires. And it's a useful reminder for anyone in hospitality experimenting with AI tools right now: don't give in to the temptation to just rewrite the one bad output. Invest a little time to fix the underlying logic, and you’ll save time in the long run. 

The next steps for hoteliers 

If your PMS has an API, there's no need to wait for a native integration to start experimenting. Yes, it will get easier once providers build purpose-built connectors, but the door is already open. 

Cole’s advice is simple: for a week, write down the repetitive things you do. Then hand the list to an AI tool and ask it to build you a guide. Use AI to teach you how to use AI. You’ll be surprised at what you find. 

To listen to the full conversation, watch the Matt Talks Hopsitality episode wherever you enjoy your podcasts.  

Watch the episode