Roland Siebelink — Host (00:00)
Hello, everyone, and welcome back to Breakthrough AI Operators. I'm so glad you were able to make it again. This is the show where we talk with founders who made AI structural — not cosmetic, not just demos, not productivity theater — real operating model change. And today's guest is a prime example of that. He's processed more than 300 million customer service calls for brands like Under Armour, Tory Burch, and the NFL, while building the company with roughly 70 people.
‍
The part I find most interesting is how it started. He was building a college ride hailing app, and the real value turned out not to be the rides — it was the broken phone support infrastructure that was underlying the service. He took that insight and made an early bet on vertical AI before it was a category that investors were even talking about. In his own words, the reason was simple: deep industry specialization is where the real application moat lives. With that, everyone, meet this week's guest, Brian Schiff. He's the co-founder and CEO of Flip. Hello, Brian.

Brian (01:05)
What a lovely introduction. Although I can't say that I have processed 300 million phone calls myself. That sounds unbelievably tiring, but we've certainly built a product that is operating at that scale.

Roland Siebelink — Host (01:20)
Well, whether it's us or our agents doing the work these days, it's all the same, right? But thank you for being honest about it. In a way it's slightly reassuring. I'm not sure I could get through a podcast interview with somebody who's actually done 300 million customer service calls all by themselves. You'd have to be a very specific kind of personality to have survived that, I would guess. Brian, the key origin story — of course, you started with a college ride hailing app and then what did that failure reveal for you about the real problem that your operators were faced with?

Brian (02:00)
I met my co-founder a decade ago now. And you hear the "10 years for an overnight success" thing, and then it's a little bit different when you live the 10 years. We met in undergrad and really wanted to build something. Now is a crazy exciting time to be building, but even then, technology was moving really fast.

The initial spark for us was everything that was happening with the gig economy and ride sharing and Uber and Lyft. That was our first market opportunity. We jumped in. Ride sharing was still banned in upstate New York at the time, and we were able to establish contracts with these local taxi companies. We handled the technology and the marketing on these campuses and you can imagine the amount of fun that we had.
‍
It was really where we cut our teeth — what does it mean to build production software? All the different components that go into creating a business from nothing. But it wasn't gonna be a long-term opportunity. We really wanted to build something that carried more weight, more scale, more everything with it. And as we were working with these taxi companies, no matter how well the app was doing, the calls would just ring off the hook incessantly.
‍
And this was 2018 and it was right when the last era of AI was having its moment. It was Alexa and it was Google Home. We pivoted the business and started building with that version of this technology, started to figure out how do you take the technology, turn it into a product that is going to be able to ultimately deliver the business outcomes for these companies that are using it — and do it in a way that is easy to pick up, easy to implement, easy to measure the outcomes that you're getting, easy to administer it. And then how do you wrap that in a business model and in a go-to-market motion that allows you to work with all of the companies in that industry?
‍
The temptation of "I'm gonna build the AI everything for everyone" — everybody gets stuck trying to be too much for too many people because it's just the pull of big market, limitless opportunity. Why can't I do it? And when you force yourself to be hyper-focused, you then start to think about: how can I get every company that is like this company for it to be a no-brainer for them to use our product? We really went through that journey for the first couple of years post-grad in the transportation space. It's one of these — who would have thought that the taxi industry would be at the frontier of AI adoption in their customer experience? And that's the reality of how it played out. A lot of it was the pressure that they were getting in the marketplace from Uber and Lyft and ride share. And also their willingness — they were looking for technology to combat the technology that was disrupting them.

Roland Siebelink — Host (05:18)
Double-click for a moment on what you said — everyone goes through this phase of wanting to build the AI for everything and everyone. Were you inoculated against that because you were already working in the transportation industry with these taxi companies? Or did you also go through that phase where you said, my god, the possibilities are limitless, let's just build something limitless?

Brian (05:40)
There are two types of pivots. You can do a market pivot or you can do a product pivot. For us, we did a product pivot. We were already working in that market. We were already selling to those companies. And this was us saying, we want to continue selling to you, but we want to sell you something different because by working with you and learning your business and learning from you, we have found what we think is a much more powerful solution.
‍
I think a lot of credit goes to going through Cornell's eLab program — they preached the fundamentals. It was very much about: you need to be concrete on what's the problem that you're solving, who are you solving it for. And there's a lot of smoke and mirrors and marketing and all these things that are going on in the world of AI right now. But it was the fundamentals. If you want something to work for a lot of people, then you need to make it work first for a small number of people.

Roland Siebelink — Host (06:42)
I also find — and maybe this is something they talked about in the Cornell eLab as well — there are also two startup types in the early stage. There are those that have a solution that's looking for a problem, and then there are those with a problem looking for a solution.

Brian (06:58)
Our category — we do AI for customer service — which is today, in summer of 2026, one of the obvious big opportunities for AI in the enterprise.

When C-suite at large corporations are sitting down and saying, how can we apply AI in our business today to drive big impact, this is on the top three list. There's a lot of attention, a lot of activity, a lot of funding, a lot of startups. And there are so many companies that have approached this market, I would say, parroting back your language, from the technology outward rather than from the customer inward.
‍
They haven't said: there's this business that I deeply understand, I understand how it works, I understand the pains that they have, and I want to solve that problem — and this technology is going to help me solve that problem. They've instead approached it as: AI is this amazing new technology, I want to build a company with AI, and let me figure out where it can be applied. I'm going to apply it to customer service. It is very technology outward versus customer inward.

Roland Siebelink — Host (08:13)
Yeah — come listen to my agents and see them pick up a phone call without ever having thought about processes or back office systems or all that stuff.

Brian (08:23)
And here is some tooling. And I'm sure that you guys have a huge team of engineers and product people and AI experts that can go and build the agent and integrate it with your systems.
‍
Technology companies are filled with engineers, right? You are a technology company, you're building products. When you go out into all of these other industries that you're serving where this technology is ready to be used for this use case — big consumer industries — they're not full of engineering teams. Their product organization is their team that is deciding what are the shirts, or what is the formula for the supplement that we are creating. They are creating goods and services that are not software. And the idea that they just have these big teams of engineers and data scientists waiting to build an agent from scratch on top of your LLM wrapper — that's not reality. And there is nowhere that that is more true than in the taxi industry. Do you know how many engineers work at a taxi company? Zero.

Roland Siebelink — Host (09:50)
Is that something that, when you talk to founders that are maybe a little bit behind you, maybe a bit younger, is that something you impress on them? You have to start from an actual problem and not just from a technology?

Brian (10:04)
Yeah, I think that it comes up. I try and stay off my soapbox, if you will. I think wisdom is good, but you can find examples where every strategy works. There's gotta be that concept of founder-business fit — and people need to find that. There are a lot of things that we did that were cutting against the grain when we were building this business that have now, in the balance of time, become the de facto way that AI companies are built. And everybody was telling us, why are you doing that?

Roland Siebelink — Host (10:56)
Do you have an example? Something you're comfortable mentioning?

Brian (10:59)
Roland, who would I be if I showed up and I didn't have an example?

Roland Siebelink — Host (11:03)
I love that.

Brian (11:05)
A super popular business model with AI startups today is usage- or performance-based billing. Rather than the classic SaaS charging for seats, I'm gonna charge you for the amount of phone calls that I automate.
‍
(11:21)
We started doing that from the beginning. And it was first principles that got us there. We were showing up with a new technology to skeptical buyers, but if the technology worked, it was going to be transformative for them. And so what we needed to do was remove all of the risk of trying it and shoulder all of the burden of it working. And so we showed up and we said: no money up front, two-week free trial. We are already integrated with your tech stack. We're already working with some other people that you know in your industry. And we are only gonna charge you for the phone calls that work. We're only gonna charge you for the queries that we solve. And for everything else, we're gonna escalate it.
‍
(12:12)
And for us, making decisions from the customer back — really thinking, what is the right thing for the customer that we are talking to at this moment in time in the business — that was the obvious way to price this. But we would then go and talk to investors or seasoned operators, and they were like, it's like the payment processing companies — that was the closest comparable to a usage-based billing model. And the payment processing companies never got SaaS multiples; they would get valued at 3X or whatever it was versus SaaS at 10 or 12X. Everybody that was smart and wise and had been around the block was like, why are you guys doing this? That's clearly not the right way. You're not even a SaaS company. But for us, we were just operating on what makes sense.

Roland Siebelink — Host (13:05)
I love that. Were there some other first principles that led you to different directions than what was conventional? Because you're really talking about counter-conventional thinking, right?

Brian (13:17)
We from the beginning took an industry-specialized approach to this category, the category of AI customer support. The de facto model that most people take is: we're gonna take the models, we're gonna package in all the core technology components, we're gonna create a UI that allows any company to build their agent from scratch on top of the platform. And then we're gonna go and sell it to companies across every industry — a retailer one day, a healthcare company the day after that, a bank, an airline, a utility company.

We took the opposite approach. We said, we're gonna build for one industry.

Rather than just providing tooling and leaving it to every customer to build their agent themselves, we said we're gonna build the best AI agent for a retailer. Imagine a customer service rep that has answered 300 million phone calls on behalf of the best brands in retail — that you as another retail brand have the opportunity to hire and have work for your organization answering all of your phone calls. We built all of that intelligence, all of the know-how around what are the nuances and the edge cases that you need to account for when canceling an order, when issuing a return label, when answering product questions, managing a subscription, scheduling appointments.

Rather than leaving it to every company to build all of these workflows for all of their call topics with all of the edge cases — leaving it to them to figure it out, to go through six months to launch a first version and then see all the issues with version one, need to get it to version 1.5 and then version 2 and then version 2.5 — we said we are going to build the perfect agent for this industry. And then we're gonna allow every company to customize it to their specific business logic and their brand voice.

Whether you sell subscription supplements or you sell shoes or whatever you sell, the reasons why customers are reaching out are overwhelmingly the same. And the systems that your agents are using on the back end to resolve those questions are the same. Rather than forcing every company to build it themselves, we said we're gonna build it — and then we're gonna allow you to pull those strategic levers to make it exactly what it needs to be for your brand. But so much of it is the same for everybody.

Roland Siebelink — Host (16:11)
If I may, Brian — by defining your customer more narrowly upfront, you can then go so much deeper in providing the actual value. It's almost like the others, the blue ocean approach companies, abdicate their responsibility for that last mile. And you're able to build that for customers because you know it's this particular customer, these specific systems. That must provide so much higher value to customers as well.

Brian (16:40)
And what I'll tell you is it's not a last mile, it's a last 100 miles. All it takes is talking to the people running the customer service department at these organizations to realize that people don't reach out about three things and their agents don't use one system. That is not reality. Reality is much more breadth and it's much more depth. And if you are not building all of that into your platform, then you are forcing them to do it.
‍
We'll talk to somebody that has spent the last couple of years trying to build their agent on top of one of these platforms. And there is a team of 20 people provided by the vendor that is building the agent from scratch for that one particular brand. And we show up and we say, why would you want to spend the next five years building yourself 20%, 10% of what we have out of the box today? You in two years working with 20 people at their company, you've gotten three of your things launched in production and one integration. You can do an order delivery status and you can do a cancellation, but you've got 98 other things.

Roland Siebelink — Host (17:56)
And it's already proven with the back end systems that we know you use, right?

Brian (18:01)
Exactly. When you go and talk to the leaders that are implementing these solutions today, time and time again the best outcomes are from the companies working with industry-specialized solutions. And so we're doing that in transportation, we're doing it in retail, we're doing it in healthcare. That's it. We had a large telco company come to us last year and they wanted to work with us. They had heard great things from our clients. And we were like, sorry — literally, you can turn on Flip and it will not be able to do anything for you. The agent will be the dumbest agent you've ever talked to because it's not trained to handle the topics that people are calling your telco business about.
‍
Ultimately, what matters is you have a piece of technology that is capable of delivering a business outcome, and the industry-specialized players are consistently getting better and better outcomes in the marketplace. This is true in our category and it's true in other categories. Tying this all the way back — we were very cutting against the grain in the bet to be verticalized. This idea of vertical AI and even vertical SaaS was kind of a thing. You had ServiceTitan and Procore, some other companies that were doing it. But the whole first wave of AI was super generic, super horizontal. And only recently have you started to see a lot more attention and interest in this vertical AI approach.
‍
The level of scale that we were able to get to on the amount of capital that we raised is orders of magnitude more efficient than the horizontal competitors. And now you're starting to see these vertical companies really break out — because ultimately, what matters is how good of outcomes are you delivering for your clients and how hard is it for you to deliver them consistently and at scale. And it turns out doing bespoke builds for every single client is not that scalable of an approach.
‍
Part of what I love about retail is it is the only industry where they refer to companies as brands. It is so elite and it has persisted for so long, and it is so expressive of — brand is synonymous with the customer experience. They go one-to-one. And so when you look at all these industries, they all look up to retail as the premium when it comes to designing and orchestrating and how to create emotion and connection and all of those things at scale, how to treat your customers right. Which means that both they have the most to gain and they have the most to lose when they adopt this technology.
‍
When you get AI for your customer service, you are saying: I am going to purchase and implement this piece of AI that is going to have thousands of conversations with my customers every single day. Those can go amazingly well or they can go horribly wrong. We've all seen the news cycles of where things haven't gone well. And so the idea of an industry-specialized solution — an agent that is already working on behalf of Under Armour and Tory Burch and all of these marquee brands, that's been trained and optimized over so many conversations — part of it is speed to market, part of it is the performance and the business outcomes you're gonna achieve, and another huge part is just the safety factor.
‍
The idea that I don't want to trust this person on my team that is just learning how AI works to go and build something that is gonna be talking to my customers every day — there's real brand risk associated with that. And so we're able to come and say: this thing is battle tested, this thing works. This is exactly how you are going to test it, connect it, configure it, and then roll it out to make sure that this thing goes off without a hitch, and to make sure that this new superpower that you have is channeled for good and delivers the positive outcome and not the negative outcome.Roland Siebelink — Host (22:43)
You're almost providing the trust building as part of the package — or maybe even the whole change management inside the company as part of the package.

Brian (22:54)
Undoubtedly. And some of the people that have been with us for a few years have implemented this technology successfully for dozens of large, premium, household-name brands. And I tell them all the time: you don't realize how much of an expert you are and how much of an imbalance there is between everything that you know about the right way to do this successfully, and the new client, the new brand that is signing up and just starting their journey with us now. They work with us because they trust us, because we have delivered the results for 250 other brands just like them. And it is our duty — one of the things that I say to clients all the time is: we bear the burden of the outcome.
‍
It is our job from day one forever to own the ultimate delivery of the result. Not to just provide a piece of software, not to just provide venture-funded professional services, but to actually own that this thing is gonna go off without a hitch. This is what it's gonna deliver. This is what you can stand behind to your leadership team in terms of what the business outcomes are gonna be. And it's really cool for the people that work here — they are on the front lines of one of the biggest AI categories that are out there right now.
‍
I think the magic potion here is we've built our business on the idea of having a reference list as long as our client list. We know that every single client that we bring on and that we take care of — when there's a million vendors and a high-stakes decision — everybody's looking for their peers to tell them who to work with and how to cross the chasm. And so trust is the ultimate currency. And we are able to achieve that by, one, the product outcomes that we deliver, and two, the experience that our team delivers. We need to be the most impactful tool in their tech stack and their favorite vendor to work with. Either one of those on its own is not enough. It has to be both of them.

Roland Siebelink — Host (25:01)
Excellent. I love how you raise the bar for your team that way — setting really high standards and that's gonna drive your success. Can you talk a little bit about how you expanded into other verticals? Because transportation was first and I believe retail came next, and now healthcare most recently. That goes a little bit against what you said before about wanting to really deeply own one vertical. But now you're still expanding. How are you striking that balance?

Brian (25:42)
The category demands an industry-specialized product. There is not product market fit without the industry-specialized approach. You have to take the time to build the industry-specialized solution. You have to build everything into the platform. Our teams are built by industry — when you talk to our customer success team, the person that you're talking to only works with clients in that industry. When you talk to a salesperson, they're only working in that industry.
‍
We firmly believe that this category is gonna produce a Salesforce-sized outcome. This is one of the really, really important categories of enterprise software. And if you can find a way to both do it right — meaning do it industry-specialized — and then replicate that in a few different places, you can really build something special.

Roland Siebelink — Host (26:54)
You guys did tremendous growth over the years, raised good money, signed up so many customers. What would you say were some of the key breakthroughs that you encountered — or that led to a new chapter altogether or a new way of thinking about the company?

Brian (27:14)
One of those old idioms is strong convictions, loosely held. And that captures a lot of the right mindset. You need to both believe things that nobody else yet sees or believes, and run straight through all the doubters. And then at the same time, really be a student of the market and a student of your own business, and honest around where are things working and where are things not working.
‍
And when it makes sense, have the humility and the courage to drop what you're doing and go in a different direction. One of the things that I tell everybody that works here, and new employees that are joining, is: everything in this company is written in pencil. It's in pencil, not pen. And it's your job to come in and pick up the pencil and go and make it better.

Roland Siebelink — Host (28:11)
We're in the last minutes of the recording. I always am interested in the person behind the startup. Where'd you grow up? What was your childhood like, your teenage years? And was there anything that would have predicted that you'd be such a successful founder-CEO one day?

Brian (28:38)
Good question. I grew up in the New York area. In so many ways, just a normal kid. Loved sports. It's funny — I never thought that I was an adrenaline junkie when I was growing up. And now I look at my life and maybe there are some threads of that going on. But I think that there was a competitiveness, there was a team orientation, there was a curiosity, there was a self-belief. I was a hockey player.

Roland Siebelink — Host (28:59)
Hockey player — okay, cool. Yeah, that's definitely the competitiveness, the aggression, but also the team spirit that you learn, right?

Brian (29:25)
Yeah, it's funny. I was talking to one of our board members the other night and he was like, every once in a while when I'm talking to you, I need to remember that I'm talking to a hockey player. I don't really know what that means, but if it allows us to communicate more effectively, then I'm all for it.

Roland Siebelink — Host (29:34)
We always say feedback says as much about the person giving it as the person receiving it. Was there somebody in your childhood or teenage years that really inspired you around entrepreneurialism, taking control of your own life — the kind of person that saw a spark in you earlier than you did?

Brian (29:47)
Yeah. I think we were raised in a "you can do anything if you put your mind to it and you stick with it" kind of environment. And we were always encouraged — if there was something that we were passionate about, lean in and go for it and see how far you can take it. And I think that gives you an appreciation for the roller coaster dynamics of it and the twisting and the turning and all of those things. But I showed up to college and "startup" was not a word in my vocabulary.

(30:41)
All of that came from Sam. He grew up on the West Coast, in a family that had been in and around technology and startups. And he had been building and playing with things even before we met. And when I met him it was one of those light bulb moments — one of those, I now know what I want to spend my time on.

Roland Siebelink — Host (31:10)
That's so powerful when you know that for your life. That sounds like we might spend another podcast on just that topic at some point in time. And I do love repeat guests, so when your new product comes out or you're entering a new vertical, I'd love for you to come back and tell us all about it. But for now, Brian, thank you so much for this episode. And where can people find out more about you or maybe be in touch with you? And if there's something on the website, is there a particular thing that they should download or read if they want to hear more about you?

Brian (31:42)
Yeah, we're more LinkedIn than X. But you can certainly find us online — find me on LinkedIn.

Roland Siebelink — Host (31:52)
And if someone wants to be in touch with Brian — you know me, you don't know Brian yet — I'm happy to always provide an introduction. With that, thank you so much, Brian. This has been an amazing interview. Love to have you on the show and see all the success that Flip is achieving under your leadership.
‍
And for the audience, we will have another founder — maybe not quite as amazing as Brian, but still a good breakthrough AI operator — on the show again next week, so stay tuned.
‍