NYC Tourist Business Puts 3,217 Customer Conversations in One Inbox With an AI Receptionist
A two-location tourist business in NYC. Every message now lands in one inbox and gets answered 24/7, wherever the customer writes from: Instagram, Facebook, TikTok, WhatsApp, email, website chat or a phone call. When the AI doesn't know, it hands off to staff instead of guessing.
At a glance · As of Oct 2026, from their CRM
3,217
customer conversations in one inbox
The Problem
Picture a Saturday in July.
The counter is checking in a family of four, and the phone rings. It rings about 80 times on an average day, and most calls are the same simple questions: how do I get there, what are the rules, when can I come in. Meanwhile a text asks how much an e-scooter is for an hour, a DM comes in on Instagram, and someone writes on WhatsApp. Each one lands in a different app.
And the answers didn't agree. One page of the website said $30 for that scooter, another said $25. A caller asked about luggage storage, the website never mentioned it, so the answer was no and they went somewhere else.
Now multiply that by two busy locations, in English and Spanish.
Then the first version of the AI went live and made it worse. It learned from the website, so it repeated the website's mistakes. It quoted the wrong scooter price, said there was no luggage storage, and told a parent their 12-year-old couldn't ride. It even said there was no location near Battery Park City, which is five minutes from their Brooklyn Bridge store.
What the Team Does Now
- Hand over the bike.
- Take it back.
- Pick up the conversations the AI hands over.
What Happens on Its Own
Wherever the customer writes from
- Instagram, Facebook, TikTok, WhatsApp, email, website chat and phone calls all land in one inbox.
- Every message gets an answer in seconds, 24/7, including the everyday questions: directions, rules, hours.
- The AI answers in the customer's language, and in Spanish automatically.
When someone wants to book
- The AI steers them to book online, where prices are lower, with the link for exactly what they asked about.
When the AI doesn't know
- It never invents a price or policy. It hands off to staff instead.
When a person needs to step in
- An upset customer, an accident, an existing reservation or a big group goes straight to the team.
- Staff get an alert with the conversation, so nobody starts from scratch.
When a call is missed
- The caller gets a text back right away, on both lines.
From booking to return
- An online booking lands in the system on its own.
- Check-in, the start of the rental and the return are each recorded without anyone typing them in.
- A rental running late is flagged automatically.
- Finished rentals close themselves out.
What Made It Work
We audited their website first and found 14 pricing and policy conflicts on their own pages. Those became one approved source of truth, and the AI answers only from that. Then we tested it the way customers would. Every AI mistake gets traced to its root cause: the page it came from gets fixed, not just the answer.