Why Most AI Gets Events Wrong

A while back, we were on a call with a VP of Sales at a multi-location events group. She was four months into using an AI tool that had one job, which was to book a call with her team.

Instead, it told a bride the venue had an outdoor ceremony space. It doesn’t. And her team had no way to correct it.

That wasn’t a bug. The tool did exactly what it was designed to do: it was fast, polite, and confident, and it was wrong. When you’re talking to someone about the most important day of their year, being wrong with confidence is a trust problem, and trust comes down to whether the thing talking to your guests actually knows your venue and cares about hospitality.

She asked us, “When are you guys going to replace this?” We took that personally.

We both came up in restaurants, hotels, and events before we ever worked in software, and we shared the story of building Cohost at Ignite this August. Here’s the version for everyone who couldn’t be in the room.

We started by listening, and found almost nobody was thrilled

Before we wrote a line of code, we spent months on calls with about 40 venues. A pattern showed up quickly. A lot of events teams already had some kind of AI touching your leads, and almost nobody was happy about it.

What surprised us is that it usually wasn’t because the AI was bad. It was because of what happened when something went wrong. A room got recommended that you’d never sell for that event, the tone didn’t sound like your team, or it kept asking a question that isn’t how you sell. You’d flag it, and the flag went into a support ticket, a form, or a roadmap at a company that has never run an event. So you submitted feedback and hoped for the best.

We asked the room at Ignite how many people had used an AI tool that got something wrong about their own venue. Plenty of hands went up.

People book the venue that takes care of them

Here’s what we believe, and it’s the constraint we build against. Nobody books an event because of a sleek AI chat. They book because they feel taken care of by your team, and AI can’t do that part.

What AI can do is the 30 emails that come before that moment: the date checks, the headcount that changed twice, and the lead that was never going to be a fit but still ate 20 minutes of your day. One venue told us they spend 80% of their day saying no. That’s the work we wanted to take off your plate, so the human part of the job gets more of you.

That’s the bar we measure everything against: does it take care of you, and does it take care of your guests.

Meet Cohost, who lives where your events already live

Cohost is the teammate who never misses an inquiry. When a new lead comes in, Cohost opens a text conversation with your guest in your venue’s voice. It asks the qualifying questions your team already asks (date, headcount, timing, and space), checks your real availability, and gets the booking ready for your review. Cohost does the repetitive work, and your team makes the call.

The reason it can do this is where it lives. Cohost works inside Tripleseat, where your discussion history, your pricing, and your menus are already sitting, in most cases for years. That means it can pick up how your team talks to guests on day one, which is a different starting point than a tool that has to be taught your venue from scratch and only knows what someone had time to type in.

What we built, because you told us to

Every feature below exists because a customer said it would help.

The biggest one came from how differently venues want to work. We originally built Cohost to run an inquiry all the way to a booking: have the conversation, convert the lead, hold the room, send the Guest Portal link, and take the card. For some venues, that was exactly right. Others told us quickly that they weren’t ready to let an AI get anywhere near a credit card, which is completely fair. So we built a second path where Cohost qualifies the lead, covering date, headcount, budget, and the rest, and then hands a fully qualified lead back to your team to close your way. Tripleseat has always bent around how you work, and we weren’t going to force everyone down one road.

From there, the improvements came in a steady stream:

• Room selector. A space might technically be open, but you hold it for catering, or walkthroughs, or you simply don’t sell it for a 20-person dinner. Cohost used to offer whatever was available. Now it only offers what you’d actually offer.
• Real-time lead updates. When a guest says the headcount is actually 100, not 80, the lead updates immediately. If their date isn’t available, Cohost asks whether they’re flexible and checks what else you have, right there in the thread, so there’s no digging through messages and cleaning things up later.
• Lead form selection. Many venues run different forms for different occasions. You can now choose exactly which forms Cohost answers, so not every inquiry has to run through it.
• Cleaner links. Long, slightly sketchy-looking links are now shortened, so what lands in your guest’s text looks legitimate and looks like it came from you. Because it did.

There’s also a second kind of control that doesn’t need us at all. Everything above took an engineer and a release. This one doesn’t. You can write a rule in your own settings, in plain English, and Cohost follows it in the very next conversation. Not next sprint. As soon as you click save.

Meet Glenda, and the rule she wrote herself

Glenda is the Director of Sales & Catering at Crown Block, a restaurant with 360-degree views high above Dallas that does everything from intimate VIP dinners to large corporate events. With two people on her team, she could use an extra set of hands on the leads coming in the door, so she became one of our design partners.

Design partner has a specific meaning for us. It means Glenda has a direct line to our team, and when she tells us the product should work differently, it changes. She isn’t evaluating a finished product. She’s helping decide what it becomes.

She named her Cohost Bailey, and she asked for something simple: she wanted to see every lead Bailey had touched, not just the ones that turned into bookings. So there’s now a dashboard filter that shows every conversation Bailey has had, in one place, whenever she wants to check.

Then Glenda noticed Bailey was asking guests about budget early in the conversation, which isn’t how she likes her team to sell. She didn’t file a ticket or wait on us. She went into her settings, typed “don’t ask about budget,” and it was done, right then. Remember the feedback that goes nowhere from earlier? This is the opposite.

Part of Rachel’s job is reading the conversations these agents are having, not to babysit them, but because every thread teaches us what to build next. Recently she was at her desk reading through leads when a corporate holiday party inquiry came into Crown Block for 140 guests. She watched Bailey handle the whole thing: the date, the headcount, the menu, the food and beverage minimum, all while following the rules Glenda had set.

Then she watched Glenda convert the lead. Glenda had sent exactly one message. She stepped in once, at the end, and the event was hers.

Rachel was so excited she sent Glenda an email that was, by her own admission, mostly exclamation points. Glenda wrote back: “We were literally screaming.” She had watched how quickly that guest was moving with Bailey, decided that was her signal, and sent the proposal. The guest was touring the venue the next day, and the event was worth about $52,000.

That’s what a design partnership looks like when it’s working. Glenda shaped the product, and the product paid her back.

What’s coming next

None of the following is live yet, but this list comes from what venues keep asking for, and it grows every week.

Email is first. Almost everything Cohost does over text maps cleanly onto email using templates you already have, and some leads and venues are simply better reached that way. Next is recognizing returning guests. Right now Cohost meets everyone like it’s the first time, and remembering someone who’s booked with you before, and referencing their history, is a big part of real hospitality. After that is multi-location and cross-selling. Cohost works at the single-location level today, and many of you run more than one, so we want it to work across a portfolio and recognize when a lead is actually a better fit for a different location.

It’s a who-built-it problem

Back to that VP of Sales. The tool she was using never had much of a chance, because the thing texting her guests wasn’t built by anyone who has run an event. That isn’t really an AI problem. It’s a who-built-it problem.

So we don’t think the answer is less AI. We think it’s AI built with hospitality people in the room the whole time, people like Glenda, and hopefully people like you.

A co-written blog by Tess Gostfrand and Rachel Daly

About Tripleseat 

Tripleseat is the leading Event Management Software for venues, built for hospitality by hospitality. Trusted by over 20,000 venues, Tripleseat brings every lead, booking, and detail into one clear view, helping event teams spend less time chasing updates and more time delivering the moments that matter. Learn more at www.tripleseat.com.