Fixing the Frankenstack: How to Use AI to Integrate Your Travel Tech Systems

 In Episode 3 of the AI Compass podcast, Jason Halberstadt and Jason Elkins dive into one of the biggest challenges facing travel companies today: the 'Frankenstack'.
 
 💡 After interviewing 40+ travel company owners and analyzing 60+ AI conversations, one insight stood out — companies are drowning in disconnected systems. Booking engines, WhatsApp, email, CRMs, payment tools… all working separately.
 
 Learn how AI can bring these pieces together, create a unified view of your travelers, and even predict what clients need before they ask. Discover what a Model Context Protocol (MCP) is, how reverse-engineering your best client journeys can surface blind spots, and how sentiment analysis can improve your customer experience.
 
 Whether you're dealing with data chaos or just getting started, this episode offers practical guidance for turning your tech mess into a smart, AI-driven travel ecosystem.
 

Book a FREE AI strategy call with host Jason Halberstadt at MyTrip.AI.

Transcript

Jason Halberstadt (00:01)
Welcome to the AI Compass podcast for travel leaders. I'm Jason Halberstadt.

Jason Elkins (00:06)
Jason Elkins, welcome back.

Jason Halberstadt (00:08)
Great, today we're gonna talk about a bit of a unique study that we've recently done. Over this last year, I've been talking with probably about 40 or so different CEOs and owners of travel companies about AI, about what's happening with their company and AI. And also we've had about 60 or so...

companies that have come in and talked with our AI assistant that essentially holds a conversation as well about how people, ⁓ owners of travel companies and executives are using AI and the different challenges that they see ahead. so really, so what I did then is I took the transcripts from all those meetings and the chats, put them into one big chat session in Google Gemini.

which can handle almost like a million words. And then put that into a chat window and started asking questions about it. Like what are the most common problems that travel companies have today that AI may be able to solve? And so actually the number one that AI identified, the number one problem is, ⁓ actually Jim and I came up with this term, it's called Frankenstack.

Jason Elkins (01:35)
you

Jason Halberstadt (01:35)
So

what is a Frankenstack? I don't know, Jason, you want to chime in or what?

Jason Elkins (01:41)
So, so Frank

said, we were chatting about this before and I said, geez, Jason, did you come up with that term? And you said, well, no, Google Gemini did. ⁓ So that was nice of you to give it credit. But I thought it was great because I was like, what is this? What does this mean? And I think you were describing it as like.

Jason Halberstadt (01:52)
You

Jason Elkins (01:58)
someone you know maybe starts their tour operation company and someone comes along and sells them a website. Well you got to have a website so maybe they get a i don't know a Wix website or a Weebly website or whatever and then someone else comes along well you need a reservation system so here's the one we sell and well you need some email marketing you need some this you need some that and before you know it you've got you know 15 different monthly payments that you're that you're sending off for these different services and then someone comes along and says

we need to take Zapier to make it all work together and it's like it just you know putting all these pieces parts together and I know you've been dealing with this for a long time and we were discussing how maybe someone comes to you and says this is we've got this here this there this that and you're like well maybe we just need to start over from scratch so that's what franken franken stack is right

Jason Halberstadt (02:47)
Yeah,

yeah, yeah, absolutely. And actually from analyzing all of these conversations, actually the conclusion was that either people have this Frankenstack that's just kind of put together over time, or a lot of times people are just starting over. Or I mean starting from zero. A lot of companies, or actually potential companies, even...

people just thinking about starting a travel agency or tour operator have thought, okay, if I'm gonna do this, I'm gonna do it from the beginning, starting with AI. But most of the companies have this problem of the Frankenstack. And so it causes all kinds of problems. mean, just think of a typical ⁓ sale of a tour package, wherever we have...

you know, requests coming in from the website. You have to do follow up with email, maybe by WhatsApp. You have to go into your itinerary builder to make them itinerary. You have to copy and paste that into the email. And so...

Jason Elkins (04:00)
they go to PayPal

to actually pay the payment and yeah yeah

Jason Halberstadt (04:03)
Yeah.

Yeah. Yeah. And so you've got all these different systems where your team is essentially doing busy work, you doing copy paste from this system to that system and keeping track of everything that's going on because everything is desperate and desperate in different systems, different databases. And so...

Really, AI actually has the potential to be a solution to the Frankenstein.

Jason Elkins (04:40)
Well, I'm super excited to get into that because all of these different systems are saying AI. WhatsApp has an AI element to it. A lot of these different systems that people have in their Frankenstack are talking about AI. So.

Jason Halberstadt (04:58)
Yeah.

Jason Elkins (05:00)
I'm excited to hear you share with us, how do we use AI without just thinking, well, yeah, I am using AI. Every system I have is using some sort of AI, but how do we use AI to bring it together? it's ⁓ more, I don't know, what was the word you just used? ⁓

Jason Halberstadt (05:18)
Oops. Yeah. So, yeah. So AI has this amazing ⁓ ability to ⁓ say, read through a story. Yeah. For example, an interaction of ⁓ a customer and your sales team or customer and AI. And understand...

Jason Elkins (05:18)
Just bring it all together, right? Bring it all together for us.

Jason Halberstadt (05:46)
the needs of that person. And if the AI is connected, to your product catalog, then it can actually pull the trips that that person may be interested in from doing a conversational interview. So, hi, I'm Bob from Boston, and I want to go to Columbia for a week. And so,

If you have this entire story of the history of the client in one place, you can say, AI, here's the story. AI is really amazing at predicting the next token. What are the next things that the person ⁓ may want to do that they need to do in order to purchase your product or to get ready for their trip, et cetera. So ⁓ how do you get all this information from?

WhatsApp from your booking system, from your email, from your SMS. How do you get this all into one place? yeah, that's actually, I think for most companies, that is the biggest challenge of implementing, especially kind of medium and larger size companies. If you don't have your data in order, if AI doesn't have access to the story,

that they can put together from all these different pieces, then it's not going to work very

And so, yeah, that is the major challenge. And the way that you can do it using AI, it used to be really, really expensive and you have to have data consultants and chief information officers involved in getting data in order. But nowadays with AI, there are a lot of techniques and even some new standards.

one that came out recently is called the MCP or the Model Context Protocol. and so what what Model Context Protocol is it's like a standard sorry it's getting a little bit techy here but it's a it tells ⁓ different systems of how they should talk with each other. Yeah so you could have your booking system over say over here saying hey I can take reservations I can cancel reservations I can change reservations in

here's the list of things that I can do and kind of instructions of how to make the integration. ⁓ I know that's kind of really technical and maybe a complicated thing.

Jason Elkins (08:23)
Well, is this,

yeah, I'm curious, because we've got another term called the APIs, like to get systems to work together. So an API, you could explain it much better than I could, I'm sure. But what I'm hearing, or what I'm trying to boil this down to, is if each system has kind of an owner's manual that can be

Jason Halberstadt (08:33)
Yes. Yes. ⁓

Mm-hmm.

Jason Elkins (08:48)
the other systems, it's written in the same language. So if you've got 10 systems out there and they all have their own owners of manual written in the same language, then system number seven can go to system number three and say, ⁓ I see what you do. This is what I do. Let's do this together. ⁓ is that it? All right, cool.

Jason Halberstadt (09:04)
Yeah, yeah, you got it. That's exactly it. That's exactly it. Yeah,

and so you mentioned APIs. And so that is technically the way that systems interchange information. But yeah, these new AIs, yeah, like you said, we have an instruction manual of here's how you make a booking, here's how you cancel a booking, technically.

These are API calls with code.

Jason Elkins (09:38)
Yeah, all right. So if you've got these different systems, like you mentioned, you mentioned it used to be very expensive because, you know, it was a major, you know, their database administrator was a job, right? I don't know if that's what you're talking about, but sounds like it. So if you're a database administrator, your job is to make sure that all the information that you have available on your different clients, your vendors, all that stuff is in one place.

Jason Halberstadt (09:40)
That's it.

Jason Elkins (10:07)
It's not getting duplicated or deleted or whatever. But now, what I hear you saying is with AI, that got easier, but it's still a challenge. So how do we help the person that has five or six different platforms that aren't really talking to each other?

Jason Halberstadt (10:19)
Yeah.

Yeah.

Yeah, well, ⁓ soon we'll be launching the next version of our platform, which is actually designed to do just that. Yeah. Yeah. How convenient.

Jason Elkins (10:36)
wow, it's like we do. How convenient. I did not know you were going to answer that way. But it makes

sense. That's what we're here to help. what is that? should, whether it's your platform or a platform in general, how does that work without going into too much tech jargon?

Jason Halberstadt (11:00)
Yeah, you have to have basically similar to like a database administrator, somebody that really knows what information, where it is in your systems, and also kind of how that all connects together in the flow of the customer journey. And so you to know that, yeah, once they have a conversation ⁓ with the chat button on the website, then it goes.

email to the salesperson, goes to the CRM and then yeah and actually the CRMs end up being quite key in this. Actually the product we're coming out with soon is a basic CRM but it also can sit beside your

Jason Elkins (11:49)
Jason, just for anybody

that's listening that's not sure, what is a CRM?

Jason Halberstadt (11:54)
Oh yes, customer relationship management system. Yeah. And so that helps. Yeah. Yeah. Sure. Yeah. It's a database where you keep all your customer information and your customer history. Yeah. And so that one obvious place to have like your central kind of history, uh, transcripts of all your customer stories is, is in the CRM.

Jason Elkins (11:59)
So it's a database, right? I mean, a CRM. I mean, I know what it is, but I want to make sure that our listeners are comfortable with it. OK.

Or it should be. That's where we run into what you were discussing where you may are having conversations on WhatsApp that's not tied to your CRM or you've got a MailChamp or an email system sending out marketing emails. And I've seen some clients even that are using their email marketing systems kind of to try and manage their client relationships and conversations so it can... Okay.

Jason Halberstadt (12:26)
It should be, yeah.

Jason Elkins (12:52)
But where I was going with this, as you said, it takes somebody kind of like a database manager. So if you've got a client that's got this scenario going on, they've got stuff kind of spread all around, and you mentioned there's a new platform coming out that you'll be able to help them, does that require an interview with a human person that kind of understands how all these things are talking and then they interact with?

AI to make this work or is AI able to go look at your whole system and figure it out?

Jason Halberstadt (13:23)
That's a great question. It is a little bit of both. But really at the very beginning, yeah, it really takes a human that understands that, yeah, customer information, this type is in the CRM, the booking information's here, their itinerary is in this other system. And so, yeah, you definitely need some human guidance at the beginning. yeah, AI can also help with that. For example, we...

have a technique for kind of reverse engineering processes where, for example, you can take your 10 successful sales or successful clients. They're happy about the pre-sales, sales during the trip and after the trip. And you can analyze the history of those conversations.

Jason Elkins (14:17)
set to

create kind of a benchmark, right? It's like these 10 went well, so let's figure out what was consistent across the board and use that as our benchmark. Is that why you say 10 that went well? Yeah.

Jason Halberstadt (14:27)
Yeah, yeah, yeah, yeah, you can

use them as examples of this is an example of a good, you know, introductory email. And this is a good example of an itinerary that was personalized for that person that got a good reaction. But it also you can also even like reverse engineer processes. And you and yeah, and so you can follow the customer through their entire journey.

and identify, okay, in this step, we needed to talk with the booking engine. And in the initial follow-up email, we needed to connect with the itinerary building system. yeah, yeah, and so you can, amazing, you can actually analyze real cases that it's written. know, an email is usually the best place. ⁓ And yeah, and from that...

You can really map out where your data is coming from and the entire process. then with that information, give your AI agents instructions to how to handle each different stage and each different system that they're interacting with.

Jason Elkins (15:39)
And I suspect it's not just, AI, here's the process. Help us make sure that the process is documented and automated, but it's also, what are we missing? I would think it was a huge part of it because I think, you know, I've had a lot of conversations with Tor operators and if you start with, okay, well, where does your typical lead come in, you know?

Jason Halberstadt (15:54)
Yeah.

Jason Elkins (16:07)
well, they come in through my website or they come in through an OTA. Okay, great. What happens when, you know, what does that look like on the website? And it might be, well, we've got a form, you know, they filled out a form on the website. Okay, then what happens? Well, it gets assigned to someone on our team. So they've, they kind of know the process, but they might be missing the part. Like the real answer when I say, okay, well, what happens when they fill out the form is hopefully they get some sort of.

automation or an email that at least goes to them and acknowledges that they filled out a form. And I'm not suggesting forms are the way to go. That's a whole different conversation. But there's steps in the process or well, what happens when they get home from the trip? Well, we send them an email with a link to TripAdvisor.

Okay, well, did you ask, did you make sure they had a good time before you did that? That might be why your TripAdvisor reviews are not so great, right? So, you know, maybe check and make sure they had a good time before you ask them for the review. But anyway, so that's kind of what we're talking, an opportunity to figure out, what are we missing in the process? And AI maybe can help you see your blind spots that you don't see.

Jason Halberstadt (16:58)
Yeah. ⁓

I'm sorry, sorry, sorry, sorry, sorry.

Yeah, yeah, and it can also do a type of quality control as well. For example, ⁓ yeah, your TripAdvisor reviews are bad. ⁓ Maybe your person that's listened to reviews didn't see if they had a good time before. Or we kind of discovered the other day in a client that ⁓ analyzing their emails, that their kind of documentation, their templates they had for their emails was kind of using

Jargon that travelers are not going to understand, know, and so you can actually yeah, you can actually use AI For even for quality control like that

Jason Elkins (17:59)
I suspect because,

you know the term I'm sure, but something about capturing the sentiment of people and messaging. if you are having, maybe you do a video Zoom consult call before the, during the booking process and the person that you're having the conversation with is completely fine with you recording it to help you make sure you get all the details right and all that stuff.

And maybe at the end of the call, the person's like, well, you know what? I need to talk to my wife. I need to talk to my husband. I'm not really sure what we're going to do. And that's end of the call. Now you've got this resource.

You know, I don't know, this may be a little off topic, but the point being that, ⁓ email correspondence with somebody, maybe they get back from the trip and they send you an email and everything was great except for, and you're like, are they happy or not? I can't tell. So they can help with all that stuff as well, right?

Jason Halberstadt (19:02)
Yeah, and it can also help with all the details that will decide whether the person's going to be happy with their trip or not. I mean, if there was like an independent WhatsApp message that says, yeah, I'm allergic to whatever, and that detail got forgotten, and you get... Right?

Jason Elkins (19:11)
Right, yeah.

It didn't end up in the CRM, so it didn't end up on the manifest that went to the supplier.

I mean, I think anybody listening to this is like, yeah, that's happened. ⁓

Jason Halberstadt (19:38)
Yeah. No, so, so AI is definitely a way, once you have your data in order where you have access to the WhatsApp and to the booking system, then yeah, to bring that and tall into one place and to have AI say, Hey, revise this history. You know, give me, give me maybe a summary of, ⁓ yeah, different things like.

allergies and medical conditions or physical limitations. And so, yeah, yeah, you can, it's an amazing tool for the quality, ensuring the quality.

Jason Elkins (20:15)
What does this look like for

the operator that's listening to this? If they're fully on board and they've got all their stuff talking and using AI, is it kind of, I don't want to put words in your mouth, but here's what I'm imagining. We know the chat bots. You guys do a lot of chat bots for your clients that are great.

But what we're talking about is not just the chatbot. The chatbot is just one avenue of communication with the client. But maybe not. could the, when I say the client, the traveler, because, ⁓ but could the tour operator actually, I don't know, here's what I'm envisioning. Tour operator goes on his own website, her own website. There's a chatbot down in the right hand corner that normally engages with visitors to the website.

But you know what, I'm the owner of the company and I'm logged in somehow, whatever, the chat bot knows it's me. So can I just jump into my website real quick and say, hey, just wanna check how many people we have going out on Friday. Or whatever. So I've got one interface that I don't have to log into 14 different places. Is that what we're talking about?

Jason Halberstadt (21:10)
Yeah.

Yes. Yes.

Yes, absolutely. Yeah, Not necessarily the, maybe it's not the same, you know, chat interface that you go in with your clients, because you probably need a username and password to make sure everything is secure. So yeah, you have a secure page. And then when you have systems connected by AI, you can literally have a conversation with them. You can ask. So, well, yeah, yeah, yeah, actually.

Jason Elkins (21:27)
Is it? All right. That's cool.

Yeah. How are things going? How are the guests doing? Right? In a sense,

yeah.

Jason Halberstadt (21:56)
All my

guests that I have on trips right now, how are they doing?

Jason Elkins (22:00)
Wouldn't, I mean, as a tour operator, any

tour operator listening to this right now, can you imagine? Because I know on the weekends we worry about.

you know, Jesus, anybody lost it? Did all the transfers happen? Did, you know, did the guides show up? And wouldn't it be cool if you had all of your stuff tied together and you could go in on Friday and say, hey, getting ready to leave for the weekend, I'm going to be away. Is, know, what's happening this weekend? Any, what are the most likely scenarios of things going wrong? And they can come back and say, you know what, everything's good. We've, we've already confirmed with that, whatever it is. And I mean, that's pretty cool that we didn't have for a long time.

Jason Halberstadt (22:38)
Absolutely. It's kind of... No, I think... Yeah, no, there's like this idea of a concierge, you know, ⁓ an AI concierge where they understand the people's needs and their wants and can actually kind of follow through...

Jason Elkins (22:39)
⁓ Am I overselling this though, Jason? Because I'm getting excited. Sounds awesome.

Jason Halberstadt (23:04)
Like the customer history we're talking about, the entire transcript. Make sure that every ⁓ T is crossed, every I is dotted, and every single detail is taken care of. ⁓ It's like somebody kind of looking over their shoulder to make sure that everything is okay at every step along the journey from the first contact until they go home and write that TripAdvisor review.

Jason Elkins (23:32)
or the email that comes in over the weekend or whatever that maybe it's a client from the just returned or is getting ready to go. the system, everybody has email. We all know how to set up email notifications and we all know how to ignore them as well. Because your inbox gets so full, but if you've got AI kind of monitoring things and it's able to tell the... ⁓

the sentiment of an email that says, know, hey, we just got back from our trip. It was pretty good, da da da da da. And the system can say, hey, Jason, either Jason or Jason, you know, this client that has just done this really big trip, luxury trip.

you know, send in an email over the weekend and this kind of, you know, it sounds like maybe you need to reach out to this person and it can flag it for you and say, okay, this person, as opposed to the other person who's like, wow, we just got back, it's amazing, I'll send you photos on Monday, you know? And one of those you need to see and one of those can probably wait till Monday.

Jason Halberstadt (24:25)
Yeah.

Yeah.

Yeah, I think one of the most compelling things about AI assistance is that they can work 24-7. People nowadays are not going to wait for your three-day weekend or even your two-day weekend for you to get back with them about a potential sale. And if there is an emergency, absolutely. mean, most companies frequently they'll have an emergency number.

if it's like a real emergency, usually for people on the ground. ⁓ But yeah, you can have an AI say, if it's after hours, then if there's something from the sentiment of the email or for the WhatsApp, if it seems like it's important or it could be considered an emergency or urgent, then ⁓ escalate that and send it to, a WhatsApp, to an agent on their personal phone so they can deal with it on the weekend.

Jason Elkins (25:34)
It's such a better guest experience as well because I think anybody who's been around this business for a while probably travels as well. And I know that when I get someplace and the key wasn't where it was supposed to be or the transfer driver didn't show up or whatever, and I've got this emergency number of the person that booked the trip for me or someone on their team, and I call them on a Friday night at one in the morning, you know, and I get a voicemail.

Don't feel real confident that it's going to get resolved. Whether that person's going to get the message two minutes from now or not, there's that doubt. And if I call and I get an AI agent that I can kind of have a conversation with and that AI agent is, OK, yep, I got it completely. I understand the situation. Here's what I'm going to do.

Jason Halberstadt (26:09)
Yeah.

Jason Elkins (26:24)
I'm going to resolve this now. I'm gonna call the operator that should have picked you up or I'm gonna notify the whatever. Just give me some sense of when I hang up this phone, what should I expect to happen next? Because that's a huge thing that's, you know, anybody that's ever been stranded. Like, well, I left a message, that's all I could do. ⁓

Jason Halberstadt (26:24)
Mm-hmm.

Yeah.

Jason Elkins (26:47)
So it's that type of stuff. What else did you want to touch on in this before we wrap up? Because I know we, I'm not sure we stayed on task completely. ⁓ What do you think? What's important for listeners to know before we go?

Jason Halberstadt (26:59)
⁓ Gosh, yeah. I guess really the first step to having AI that can help you really manage multiple aspects of your business, it's, yeah, one of the first steps is getting your data and your systems in order, yeah, so the AI can have access to it. But then after...

Jason Elkins (27:23)
So is that something somebody

needs to do before they reach out to someone like yourself? So when you say get your data in order, that could sound overwhelming to most people. Or can you help them figure, if they reach out and book a free call with you, can you at least give them an idea of what that means? Because I don't think we have enough time in an episode to cover it, even in a whole episode, but a face-to-face, because every situation is different, right?

Jason Halberstadt (27:31)
Yeah. Yeah. Yeah.

Yes.

Yeah, yeah, yeah. And most companies that have a Frankenstack that kind of needs somebody to help manage the Frankenstack, they need a tech person who is going to hopefully know where things are at. If they don't, then that's a problem. But...

Jason Elkins (28:05)
I know

a few that don't. I've filled that role a few times, so I know that there's a few that do as well. So anyway, so it sounds like if you're thinking about how do I get AI, there's a couple things you can do. Keep listening to this podcast. Click on the link in the show notes. read the blog about this and other stuff. And then maybe book a call with someone like yourself, someone on your team too.

Jason Halberstadt (28:08)
Yeah.

Yeah.

Jason Elkins (28:33)
to kind of like, okay, here's my Frankenstack. Here's my list. Jason, can you help or do I need to shoot the thing and bury it and start from scratch or whatever? Because that would be helpful.

Jason Halberstadt (28:38)
Yeah.

Yeah.

Yeah. No, no. In most cases, Frankenstacks can be fixed. But in some situations, you may want to start over in certain aspects. If your Frankenstein's too ugly and patched together, then yeah, that's always an option.

Jason Elkins (28:51)
Alright, cool.

Maybe cheaper just start over, huh?

So listeners click the link, go talk to Jason about your Frankenstack and if anybody can help you with the Frankenstack, I think it's him. He came up with the word, I'm pretty sure he did, even if he used some AI to suggest it. So, very cool.

Jason Halberstadt (29:16)
Now...

Now, Jim and I definitely gets credit for the Frankenstein title.

All right, great. Thanks a lot, Jason.

Jason Elkins (29:28)
Thanks, Jason. Everybody, we'll see you back here or you'll hear us back here same time probably next week. So look forward to having you all back. Thank you so much.