Category: Podcast Episode Transcript

Full Transcript

Nataraj (00:02.214) Hello, everyone. Welcome to Startup Project. Today on the show, have Jeff. Jeff is the co-founder of Tensorwave. With so much of AI compute spending, I think we're in an interesting time where, first time ever, we're seeing a bunch of new Neo clouds evolve around different strategies in order to provide more AI compute that we need. And Tensorwave is one such company. So today, I'm going to talk about

You know how tends to be solving the AI compute problem. Why they're exclusively working with AMD and what it takes to build in the modern day data centers and a lot more with that Jeff welcome to the show.

Jeff Tatarchuk (00:47.79) Alright, it's good to be here, man. Thanks for having me.

Nataraj (00:49.924) Yeah, so a couple of years back, if you ask me, like, you know, there were smaller cloud companies, know, Cloudflare was a smaller cloud company. They were trying to be a larger cloud company. There's DigitalOcean, which is even a much smaller cloud company. But I always wondered, why are more cloud companies not coming up? And then once, like, the whole

AI compute spending cycle started, then we've seen a lot of Neo cloud companies doing different things. Some are focused on exclusively bringing more infrastructure online, some are bringing, you know, some companies like OAS data bringing whole new architectures. I think every time we sort of have a big wave of new innovation, there's always a new architecture innovation that comes up and then we see companies built around that. And I think you are one such company as well.

So, you for those folks who have never heard about Tensorwave, can you give a little bit of introduction of, you know, what is Tensorwave, what do you guys do, and how did the whole idea of Tensorwave come together?

Jeff Tatarchuk (01:58.05) Yeah. So TensorWave is simply put a Neo cloud that only deploys AMD GPUs. And it started out. So my other co-founder, Derek and I initially had launched a FPGA cloud business about eight years ago. And we were solving one of the harder problems first. FPGAs are usually more of an edge chip that's used for really low latency use cases.

And we decided to make them available at scale in the cloud. And we were one of the largest FPGA clouds previously. so, yeah, yeah, yeah. So VMXL was the company. And then we were working with a lot of the chip providers, Xilinx at the time, and Intel who had acquired Altera and then recently kind of decoupled Altera. And once they realized how challenging FPGAs can be,

Nataraj (02:34.007) was that VM itself.

Jeff Tatarchuk (02:56.21) It's we started with that. And so we were mostly focused on video transcoding and weather modeling and not really focused on AI at the time. But it taught us what we needed to know to deploy cloud infrastructure, set up data center infrastructure, create a really easy experience as possible for the end user.

And so we were doing that for some time and we realized that, you know, as soon as the market shifted all into AI after chat GPT was launched, all the attention shifted from any resources that were going into FPGAs into GPUs. And, you know, in 2022 and 2023, there was huge demand for GPUs. Nvidia had supply shortages. You couldn't get access to anything. And we had a friend come up to us asking, saying,

And, you know, knowing that we were doing cloud, you know, we were kind of AI adjacent asking if we could help him and his portfolio companies get access to some GPUs. And we said, you know, we're not really focused on GPUs right now, but we do work with, and we had worked primarily with Xilinx and Xilinx had gotten acquired by AMD about four plus years ago at the time. And then our company

We were working with AMD. We became kind of their support internally with our FPGA cloud. They would send us all of their latest and greatest FPGAs and we'd get them, you know, we'd co-develop and work and get them debugged for them in the cloud. And so we had already kind of built and were embedded into AMD. And then when they announced their GPU offering, it made sense for us to make the shift to go all in and deploying.

their GPUs at scale. And so when our friend, my VC buddy, came up to us saying, hey, can you help us get access to some GPUs? We said, would your portfolio companies consider going after AMD? And he said something that I never forgot. And he says, if it works, we will definitely encourage them to use AMD. And the light bulb went off the next day. Frankly, we created TensorWave.

Jeff Tatarchuk (05:03.566) and called AMD and said, hey, we need a significant allocation of GPUs and we're going to go all in to be the first and best to deploy your GPUs at scale. And we were announced December, 2023 as one of AMD's official launch partners for their first kind of data center chip, the MI300X. And we've been off to the races ever since. so it started with, you know, the supply shortage and then the frustration that the customers were having around

you know, the, the pricing and the profit margin that, Jensen and video were demanding. And we kept having people coming to us saying, Hey, we're tired of giving our, you know, profit and margin to Jensen. We need another solution. And so helping bring optionality to the market. we, with, our experience with AMD and seeing, you know, Lisa Sue's vision and roadmap on where they plan on going. knew that AMD was the next best solution to,

to go all in and be that for them. And so we are the premier AMD support. If you are considering AMD, we're your best solution to make it as easy as possible to deploy their chips at scale. So that's how it got started. And we've been off to the races ever since.

Nataraj (06:21.318) Talking a little bit about the development of FPGA Cloud, what was that business like? Who were your customers? Because I think that will help understand us, how did that help in transitioning towards building.

Jeff Tatarchuk (06:37.868) Yeah, so I mean with

With FPGAs, there's a lot of significant challenges around it. There aren't compilers written, or if they are, they're not very good. And everybody was trying to solve that problem. And one of the things with an FPGA is, yes, theoretically, it can be more flexible and more efficient and performant than a GPU. But the amount of complexity it takes to squeeze out that extra performance.

we found out very quickly wasn't worth the squeeze. so even if we could get an extra 10 to 15 % performance boost, the amount of extra work you had to do wasn't worth it. And so that's one of the lessons we learned very early on as we were one of very few companies working in this space and working with a small group of ecosystem.

Nataraj (07:32.262) Are you trying to build a cloud for FPGAs or are you?

Jeff Tatarchuk (07:35.694) There was a couple, but not very many, but it was more so the ecosystem that we were working with, working with Zylinks and Altera, and there were a few other researchers that were also trying to solve this problem as well. so everybody was kind of leaning in. I'll never forget being at Intel and Intel is like, asking us the same question, like, who are your customers? Like, who's actually buying these things? And we did have a number of customers that

you know, that we're building various products with like whether I said, what, like whether modeling products and video transcoding products, a lot of simulation, products. so it, it was, it was rough because there wasn't a lot of tools available. then the developers needed, you know, cause FPGA is in order to read, you know, for those that don't know, and FPGA is a field programmable gate array. that is, you know, very, very fast and

you can, you know, reprogram an FPGA, but before you had to be an electrical engineer to go in and reprogram, you know, and bear a log or whatever. but now like the engineers that can create, you know, reprogram an FPGA and create the bitstreams necessary to do it. They're very, very expensive, and specialized. And so I ran into a lot of just challenges early on as we were, you know, putting all this together, but you know,

It taught us everything we needed to know as we were deploying the data center infrastructure to support this, building our relationships with our OEMs and dealing with building a cloud platform on top to make sure that it's as easy as possible. our initial goal was to create a cloud platform that abstracted away what was underneath the hood completely. And so they just could say, this is what we need.

and we can spin it up and give them access to it. so working on some of those problems early on is what gave us the experience to it. And it was funny when people were, you know, when we first went into doing the GPUs with AMD, everybody's like, you know, these could be very challenging. You know, the first, first batch of AMD GPUs, it be a lot of problems. And we're like, man, you don't know the kind of problems we were dealing with previously. Like the problems we're dealing with on the, on the GPU side is nothing in comparison.

Jeff Tatarchuk (10:02.784) So yeah, so building out the infrastructure, supporting customers, and then making a platform as easy for the customers to use as possible with a very complex chip was our first kind of strike at this.

Nataraj (10:17.476) And for the PGA cloud, is it similar to building a regular cloud where you're working with like different data center vendors, who are your partners in that effort?

Jeff Tatarchuk (10:31.158) Yeah, so we actually we started off co located when we first built we just needed access to power and we needed to do it quickly. So we actually got set up in Cheyenne, Wyoming. And then quickly.

Nataraj (10:44.676) Like co-located and this is for the audience, like there's already working data centers and we basically take some space in a working data center and put your stuff on infrastructure there.

Jeff Tatarchuk (10:55.288) That's right. That's right. And so, yeah, they'd already built the data center. They were managing it. All we had to do is bring our servers, deploy them. yeah, they had a team working there around the clock that would manage and maintain it. And if we had any issues, they would go in and help us with it. And so we started with that and then we were able to actually build our own, data center. And, it's where we were able to get, really creative and work on some, some new and exciting efficiencies that that made it,

even better for us to do it ourselves. We're actually able to save a lot more money and learn along the way. we, learned the whole stack from, you know, acquiring the power to building out the data center infrastructure to managing and maintaining the servers to building out the software and cloud stack to support it with a very small tiger team at the time. So, yeah.

I still have PTSD from those days, it was a great experience and learned a lot of lessons about when you start a company. There was a lot of incentives that Cheyenne Wyoming had given us to move out there. And they'd given us some grants and different things, and all of that was contingent upon us hiring people for…

you know, to fulfill those grants. And if you're in a town or a place where people don't want to live, it's very hard to recruit. And so we were trying to recruit from the coasts and in a town that didn't have access to the infrastructure or housing with amenities or resources. It became, it became a challenge. And so we thought, Hey, like, this is a place where we can, you know, be a big fish in a small pond. can grow, we can work with the city and the government and make all these different moves and they're to give us money.

But sometimes when people give you money, comes at a cost. And so you have to count that cost when you're starting a company. that was one of the things we definitely learned early on.

Nataraj (13:00.55) So in some sense you will write time, write place with the right experience, deeply embedded into the infrastructure space and how to build data centers. And you basically are in some ways position to start TensorWave if you look back at it. So let's talk about TensorWave. So you're now building data centers with

AMD GPU clusters that offer AMD GPU clusters. And top of that, you offer a cloud or platform on which I could just apply my own compute clusters, do inference training, all that stuff. Is that the right way to describe conservative?

Jeff Tatarchuk (13:46.776) That's right. So we, yeah, we, we will do the whole thing from identifying power, building the data center, retrofitting older data centers to support these GPUs and then build out the whole stack for customers that need access to compute for both training and inference and production.

Nataraj (14:07.598) What does building data center today look like? How long does it take? What are the challenges that we're seeing? How much of the challenges that you see in the news about getting access to power, getting access, like how quickly people want to get these things running? Or how much of that is high-pressure? How much of that that you see on ground?

Jeff Tatarchuk (14:31.468) No, it's a real problem. Power is the commodity that is the bottleneck right now. They can make the chips. The supply chain is fine for now. But there isn't currently enough developed critical power to deploy the GPUs. And so there are a lot of companies out there that have access to power.

that are trying to sell it, but you you need to be able to get your substations and everything built and set up and the water and all the other pieces necessary in order to, to build your data center on top. And those timelines are a lot longer, take a lot longer than, people usually anticipate. And so, yes, like getting access to power is a major concern. there isn't enough power available, I think like in order to hit the current demand over the next few years, we need like 30 gigawatts. And if you think like a,

It takes one nuclear power plant to power one gigawatt of power. you know, how fast are we putting up, you know, nuclear power plants today? It's just not fast enough. so one of the things that we focus on is, you know, we have a pipeline of opportunities for us to build and they could be, you know, completely open, you know, greenfield opportunities that allows us to, you know, build on top and do all of those things. But for us, we have to optimize for speed. We have to be able to build.

quickly and deploy quickly. have customers that want it now. They said optimize for speed. They're willing to pay a premium for being able to deploy fast because this is a race and the big AI labs need access to as much compute as they possibly can. And yeah, as I mentioned, the biggest bottleneck is getting access to power. then we're seeing today, so you can, there's a lot of stranded power out in the middle of Texas.

And we have some of these other larger projects that we're seeing that yes, you can get access to power and you can build these data centers, but getting the people out to those data centers to act, to build and support the data centers is the challenge. Cause it takes thousands of plumbers and electricians and you know, everybody else to support and build it. So if it's out in the middle of nowhere, you know, you almost have to build your own little town around it to support it. So that comes with its own individual challenges and

Jeff Tatarchuk (16:56.738) There are some towns as we're seeing that are running into permitting issues or having the towns or cities, you know, protests, having data centers in their town. They don't want, you know, AI taking over their jobs. And so there's a lot of moving parts that have to go into identifying data centers and making sure that the data centers and power and builds all happen within the proper timelines for the end user. that's,

you know, really what's most important to them. So yeah, there's a lot of moving parts and then it's the financing those things that have to go into it as well. So

Nataraj (17:33.991) Yeah, that was my next question. I mean, you guys raised 100 million in CVC. It looks like a lot of amount, but when you're thinking about constructing a lot of data centers, then it still looks like a small amount considering, I think, big tech is spending about 600 billion in 2026 on capital expenditure of building data centers. So how's, and you guys have built, I'm assuming, I think two data centers already.

that are up and running or is the second one already up and running. But you have two data centers up and running, which is quite a cost for like, how does the financing today work and how are you sharing costs?

Jeff Tatarchuk (18:19.214) Yeah, it is a lot. It's a lot of money. When we got the hundred million dollars, it just comes in the bank and right out the bank. always tell people that we're still living on ramen, even though we've raised, you know, mid nine figures currently for our data center, you know, GPU deployments. And so luckily we do have, you know, great partners when it comes to financing with our investors, our lead investor.

is Magnetar. Magnetar is the fund that took CoreWeave from Seed to IPO and backed their financing. And then they're able to help bring in some of the other larger banks to also help with the financing as well. so cost of capital is an important factor in this. We are new company. We're only two years old. And so

you know, being able to number one, like get access to customers that have a great credit rating and the better the credit rating, the better our cost of capital is. but then we're able to lower the price of our deployment for the, for the end user. so, we are in a great position with funding, to do everything we need. But again, it still comes down to, you know, lining up or playing air traffic control to line up your.

power with your customer, with the financing and making sure all of those things land and can be built and deployed at the time necessary and coordinate all of that is really the challenge behind it. one of the things you mentioned early on is that, there are a lot of people attempting to do cloud. Before it was just the hyperscalers and then a couple of others. And then all of a sudden, one of my friends at Nvidia said they have 150.

you know, Neo clouds at, you know, that are like that. I don't even feel like there are 150 AI labs, you know, in this ecosystem. And so how, how is a, you know, ecosystem with 150 Neo clouds actually able to stay alive? And I do think some people saw, you know, the, the, the, the profit opportunity to make money where people think all you need is a data center, some servers with some GPUs in them, you plug them in and you rent them back to.

Jeff Tatarchuk (20:42.606) open AI and you're fine when that's there's a lot more that goes into it that a lot of people don't take into consideration as they are as they're doing this. And so I think some people just thought as a financial arbitrage and I think those that saw it as such will find out the hard way that there's a lot more that goes into it.

Nataraj (21:07.406) I think the difference would be companies who on paper can just say, go to a company like Equinox or Equinix that builds data centers and just partner with someone like that. And then if you can arrange financing and buy GPU clusters from Nvidia, then put that on the same data center and create a cloud on top of it and rent it out. That seems like a doable version given if you're good at raising certain amounts of capital on paper.

I think there a lot of companies like that, but I have not considered them as know, clubs because there is still a technical challenge of bringing a new type of GPU into the data center and building the racks, the cooling systems, the compilers, know, NVIDIA SCUDA. Now you have to make sure that if I'm a large AI lab, I'm

running my models on both Nvidia and AMD. So that means there's a technical challenge of how do you make that happen for your customer who is running a large cluster and they already have Nvidia already in another one easy way to run on your cluster. So there's a technical challenge of building the actual rack and making it available to a customer through your cloud. think that if we count that way, I think there are not that many Neo clouds.

But I think the NeoCloud market definition is slightly different because on paper a lot of companies can look like NeoClouds which are not actually NeoClouds because there's no technical differentiation or there's no technical challenge that is being solved as a company, right?

Jeff Tatarchuk (22:46.402) No, no, it's all it is spreadsheets and financial arbitrage. and if you put all of that together, you can, you know, make it happen, but the rubber meets the road when you actually have to deploy it. So GPS can be finicky. And, as you mentioned, like there, there are a lot of, there are a lot of challenges at stake. So

We love it though, because we worked on some of the harder problems first in the FPGA world, like seeing some of the opportunities that we're able to work on with AMD to truly bring this to market is really exciting for us.

Nataraj (23:24.474) So, I mean, you obviously talk a little bit about your role. mean, you're a chief growth officer. You're trying to grow this thing. Talk a little bit about customers, right? One thing is, all the top AI labs want more and more compute. That's pretty obvious. But whenever I think of training workload, how many, there's, think, a larger market for fine tuning. I think a lot more companies are doing fine tuning with either smaller models or larger models.

But how many companies are going afterwards big clusters and how do you see that demand shaping up?

Jeff Tatarchuk (24:00.226) Yeah.

You're right. I do think at, especially at the enterprise level is where we see fine tuning is the bigger opportunity. And there are a lot of companies out there that are working specifically with enterprise where they can, you know, take one of the larger models and then fine tune it to the enterprise's specific, you know, use case or needs. And that being done at scale, I think is a significant opportunity. And still, I do feel like the enterprise is still trying to navigate the AI.

you know, landscape and how they're going to integrate and implement it into what they're doing. And so I do see a lot of pent up demand on the other side of enterprise that hasn't fully broken yet, but there are a lot of people trying to solve that. on the other front, yes. Like, I mean, you, asked how many, I, again, I don't think there's more than a hundred that are doing significant like hero training runs that need more than like a thousand GPU clusters for years at a time.

I could be wrong. Maybe there's a lot more hiding underneath the bushes, but I can't imagine there being more than that. And, you know, the primary focus is on being able to support the top 10, you know, hyperscalers and AI labs that need access to compute. And so for both training and for inference now with AMD, AMD started out optimizing their GPUs for inference and

That was their first use case and making it optimized for that was really important to hit the ground running while Nvidia was focused more on training. AMD was able to capture a lot of the inference market. Meta announced that they host their llama models on AMD. Obviously, Azure has AMD in their platform. OpenAI just announced.

Jeff Tatarchuk (25:55.906) A few months back that they're doing a significant deployment. They six gigawatts of AMD will be deployed in the future. And so we're seeing a lot more of the AI labs interested and focused and they're still their primary focuses on inference. But as of recently, we are seeing more like if you look at our 8,000 GPU cluster that we deployed of the MI325s, it was built as a training cluster.

And so Lisa Sue had given us the challenge and the mandate to create the most performant AMD training cluster. And we did in record time. And so we've been able to focus on that. And so from my perspective as the chief growth officer, as I like to joke, the chief GPU officer, as I'm meeting with the labs that need access to compute, they have a lot of people banging down their door.

The amount of customers I have that are, you know, that announced that they've raised 50 to over a hundred million dollars. They're like, man, we're getting so many steak dinners from all of these AI clouds and getting flown on private jets all over the place. And they have a lot to choose from when it comes to the Nvidia world. But if they want to make a bet on helping to diversify and democratize, you know, what they're using, they have to look for an AMD solution. And, you know, we are the best with that.

When we first started, mean, was, was the challenge of people saying, I had no idea. And he did that. And so it was coming up with as many different creative ways to get the attention of the market, to make sure that they knew that AMD yes, has a GPU. Yes. It, you know, can do inference and yes, it also can do training and giving people options to try it. And so I'll never forget when we first started the company, like

three months in, we started in December and then GTC in 2024 was in March. And we had just raised our first like 40 million bucks and we decided to get a one of those LED trucks and we would surround the San Jose Convention Center during GTC. And we had, you know, a comparison of the H100 and the MI300 on the back and showcoasing all the different specs and showing that the MI300X definitely is better. And at the end, there was this robot.

Jeff Tatarchuk (28:21.07) that had a red pill in its hand and it went like this to the audience, to the thing. And everybody loved it. It blew people's minds. knew exactly what we were trying to communicate. And people were taking selfies in front of it. And everybody at a lot of the big companies that were at GTC at the time were like, yeah, you guys were all in our Slack channel as a way of getting the attention. So that people just needed to know, right? It was kind of a grassroots guerrilla marketing.

Nataraj (28:27.088) matrix.

Jeff Tatarchuk (28:49.25) that needed to be done to just kind of shake people like, there's a new player that is viable on the market and we should consider it. And so we got a ton of leads coming in and how we took, like a lot of people are just curious. know, initially we had a lot of people coming who had worked for all the national labs, which do, you know, lot of those supercomputers are built using AMD. And so they were already familiar with AMD. It was kind of our initial audience that was coming over. then a lot of, we, have some.

clients that have never bought an Nvidia GPU, like intentionally, just because they love the, they love AMD, they love what they stand for and they're committed to it. And so we still see that kind of transfer over from the consumer side and to those that are now developing in the AI world who want only AMD. And then it's, it's then it's, you know, we still have to go out and prove ourselves that we, know, that it does work. It's just as good, if not more efficient than an Nvidia GPU.

And we can show that we are the best at supporting them at that process of scale. So we typically will bring in a customer, we'll analyze what are they doing, what's their use case, what frameworks are they building on? And then we will then go and validate to make sure that everything that they're using, our internal AIML team will validate what they're doing and then make sure that there aren't any gotchas or bugs or any issues that they're gonna have.

run it and then once we validated it, and sometimes there are issues that we have to go back to AMD, we have to go back to one of the frameworks and fix something before we let the prospect on. And then once we have all that ironed out, we then get them on a POC to let them, you know, take a test drive, sit behind the wheel for themselves and show them that yes, it does truly work and it's just as efficient and better than Nvidia in a lot of ways.

Nataraj (30:44.294) Are there any specific architecture advantages people get by using AMD? Like is there a differentiation that AMD does better than Nvidia?

Jeff Tatarchuk (30:44.45) But even.

Jeff Tatarchuk (30:54.786) Yeah. And so like right now AMD still has an advantage and landed with the advantage of having more, more memory. They have more VRAM significantly than, the, than Nvidia does. And so if you need to host some of the larger models, like for instance, if you're hosting like a 70B model, you need, you know, two.

GPUs in order to do it on an H100 that has only 80 gigabytes of VRAM versus the MI300X that has 192 gigabytes of VRAMs. You're able to host more without having to kind of split it up over across more GPUs effectively. And then AMD has their chiplet architecture that Lisa bet on early on. And as a result of the chiplet architecture,

Yeah, it's starting to pay off showing that now they can break a chip down into more pieces and they could take advantage of a chip even more so than what you can do on Nvidia.

Nataraj (32:07.142) One of the biggest advantages that NVIDIA has is obviously CUDA, it provides libraries, compilers, and debugging tools for GPUs and sort of like their proprietary system that will run on GPUs, NVIDIA GPUs, designed for NVIDIA GPUs, and that has been long argued as one of the biggest sort of NVIDIA mode out there. How does that affect customers?

Jeff Tatarchuk (32:27.15) Mm-hmm.

Nataraj (32:34.096) for trying to run out of AMD. Like, the development team might be already familiar with, you know, building on top of Nvidia. And this is sort of like a, in some sense, it's not a unique problem. I think in some sense, like, when AWS, Azure, and GCP have, like, if your organization has AWS as a strength, then you hire more AWS people. That's sort of like compounds, and that's one of the reasons why AWS is so popular in startups, because the early talent knows AWS in every startup.

sort of builds on AWS, right? They have the small company advantages. So, talking about those challenges of having CUDA as a big model, how does that change, what are you guys building? Because now you're sort of like one of the earliest clouds that is building AMD clusters. How does that look like for AMD?

Jeff Tatarchuk (33:21.17) Yeah, so that was the you know, the initial kind of gut responses, you know, how, is the software and what we found is yes, there are some folk that have built on some of the CUDA specific libraries that are proprietary and if they were to have to switch, you know, if they were doing like Kublaz or CDNN or

you know, kudy and then or something like that, they would have to take some time porting it back from that to AMD. so, aim, Nvidia has CUDA and then AMD has Raco and, but what it simply means is like most, what we're finding is most AI engineers are, using, you know, PI torch, TensorFlow or Jacks. And if you are building with any of those frameworks,

You can port your code over from CUDA to or NVIDIA to AMD seamlessly. This was actually one of the things that we used to raise our first initial money was there was an article that was done by Databricks and the Mosaic ML team at the time with Naveen and Abhi over there. They were simply showcasing that they did it on an MI250 that they could showcase that you can.

you know, take your code from Nvidia on CUDA and run it straight out of the box on AMD and it works. And that was when the light bulb went off for us. This was in, I believe like late 2022, 2023 or sorry, 2023. Um, and it, it revolutionized everything. Like, we didn't even know it was possible to do that. And so, um, yeah, it's, it is, um, you know, uh, yes, there are challenges, but

Every day, AMD has done a lot to catch up on the software side as things are becoming more and more efficient. And one of things that we've done at Tensorwave is me and my team, launched a summit called Beyond CUDA. And so the first year we did the LED truck, the second year we did Beyond CUDA during GTC. We did it the Monday of GTC.

Jeff Tatarchuk (35:39.534) And we brought together the best researchers, founders and engineers who were building things outside of the CUDA ecosystem to showcase what they were doing and that it was actually possible and easy and just as performant to do. And we had about 400 show up and it had so much energy and excitement around it that it's continued to, you know, that the momentum has continued to this day. so,

So continuing to see, know, AMD's done a lot to support all of the other like Inference Engine frameworks from Triton to SG Lang to VLLM and the others. And so there's a lot of people working on some really cool projects that are trying to create kind of heterogeneous, you know, ability to switch from a TPU to an Nvidia GPU to an AMD GPU or whatever. for us, you know,

One of my things is I want to build an ecosystem of people that are doing these sort of things, solving these sort of problems and work on doing them together. And so we just had a customer the other day as we were telling them, you know, I telling them about, you know, cause we usually hear that it's the number one people think people say is about the, the kuda moat. And this particular customer said, you only need kuda if you don't know how to, if you don't understand, you know, how to really, you know, do low level development on the GPU. And for them, like kuda

they don't need it at all. Like they, they know what to do with the hardware and to get it to do what it needs to do. So, one of the original guys that developed CUDA, name is Greg De Almos. He's on our team. he's worked for Nvidia and Intel and AMD and did his own startup that was built on AMD GPUs. And now, you know, his main focus is on this and he now has this, this

Nataraj (37:33.734) think he's the founder of the open source scalar AI, right?

Jeff Tatarchuk (37:37.358) Yeah, that's right. Exactly. He's, he's great. And so Scalar LM is a unified training and inference stack that he's been developing. one of the quips that he says is that CUDA was always built with the purpose of going beyond CUDA and, know, growing and going beyond the, the ecosystem. so, that's one of the things that we partner very closely with AMD to.

support the projects that are out there to help build the ecosystem and strengthen it so that it is more brust, more robust and that people have the resources they need to be supported. yes, CUDA is no longer remote. AMD works out of the box. And yeah, if you think otherwise, we'd love to let you try it for yourself and we'll show you.

Nataraj (38:32.27) And so that means anyone like OpenAI can just easily change or quickly adapt and deploy that training cluster on any AMD based cloud essentially.

Jeff Tatarchuk (38:34.862) Thank

Jeff Tatarchuk (38:45.344) Exactly. Yep, exactly.

Nataraj (38:47.789) How does success for TensorFlow look like, know, three years down the line, you know, how does TensorFlow being successful will look like?

Jeff Tatarchuk (38:59.852) Yeah, I guess I'll start with me. What, what success would look like from my perspective is that the customer has viable options for them to buy compute and they don't have to be dictated by the demands of Jensen or Nvidia and that there is a true, you know, competitive market available for compute and from tensor waves.

perspective, we are able to play a significant part in that by providing the best optionality and providing the best AMD GPUs for the market to be able to consider those options. so for us, we want to make sure that we have a resilient, secure, performant cloud that they can rely on. We think of ourselves as the AI utility company. We want to make it just as

You know, just as secure and just as you're just as confident as you flipping a switch in your house. You're just as confident in your cluster running behind the scenes as you're doing your training and then you're running your inference and production. You are just as confident in that. And so for me, I would, I would see success is that, you know, our customers are happy because the compute decision that they have to make is the largest, you know, purchase they're going to make.

as a company where they're spending hundreds of millions of dollars, if not billions of dollars on these GPUs. And if they have to spend a lot of time messing with them and fixing them when a GPU dies and when there's downtime and working on all of the extra stuff, they're wasting time that they should be focused on their customers, their product, and solving the problems that they should be focused on. So I would consider Tensorwave a success.

If we've completely abstracted that away and we have a great resilient product that is able to provide their AI training and inference needs at scale. And yeah, we are the ultimate go-to for that on the AMD side.

Nataraj (41:12.102) Yeah, I think that's a good note to end the conversation. I think this has been a very fascinating conversation. And thanks for coming on the show.

Jeff Tatarchuk (41:19.17) Hey, thanks for having me.

Full Transcript

Nataraj (00:03.73) Hello everyone, welcome to Startup Project. Today we have an incredible guest who's not only built a multi-billion dollar company, but also fundamentally reshaped how businesses will connect to their customers. My guest is Andrew Bulecki, co-founder and CEO of Clavio, an AI for CRM platform that has become indispensable for B2C brands worldwide. Whether it's Liquid Death or Skims or Mattel.

all use Klaviyo to stay in touch with their customers and market to their customers. Klaviyo initially started as an email marketing company in 2013 and later integrated advanced marketing attribution, optimization, SMS marketing, and recently cutting-edge AI capabilities with the launch of Klaviyo AI. Klaviyo's impact has been undeniable. They are a public company worth around $9 billion, about 180,000 brands globally use them.

as of Q3 2025, that is an impressive $300 million in revenue and expected to reach $1.2 billion in revenue for this fiscal year and have impressive 32 % year-over-year growth rate. So we'll talk about Andrew's entrepreneur journey, the decisions that change gave his growth trajectory, his perspectives on the evolving landscape of data and AI.

and how they're approaching building products using AI. So get ready for an interesting and high value conversation. With that, Andrew, welcome to the show.

Andrew BIalecki (01:34.328) Yeah, thanks for having me.

Nataraj (01:36.05) So I want to set the stage for the audience. Anyone who is in e-commerce or anyone who's selling something on the internet, I think is familiar with Klaviyo. But if the audience is not familiar with Klaviyo, can you just give a brief description of what Klaviyo does and what are the products that you have?

Andrew BIalecki (01:55.182) Yeah, well, that was a good description upfront. man, so we started Clivia 13 years ago. And our dream was whether you were an entrepreneur just starting out, or you're an iconic, know, multinational business, we felt like you ought to be able to treat every single one of your customers like they were the most important only customer in the world. And, you know, we found that like a lot of businesses

when they really want deliver an exceptional experience, like just a stunning customer experience, they really relied on people to do that. And our feeling was, like, well, in the future, it's going to be less, you know, like white glove through, you know, you or I getting on the phone with somebody. And it's going be much more delivered through software. I think that's probably a take that's, I don't know, guess feels like it's aging pretty well, given all the stuff that's, you know, happening with AI and LLMs these days. But anyways, we wanted to make that

possible. so we actually, we actually started out as a, you know, as a database business. You know, because we felt like, if you're going to do that, then you're gonna have to replicate the way a human thinks and stores information. And then we actually, after we built that kind of brain, this kind of like, you know, very low latency, you know, database, but that could handle some like deep analytical queries.

You know, we gave that to a bunch of customers and they, you know, they said, okay, this is awesome. And we asked them like what they were hooking it up to or what they were doing with the results. And a lot of them said, Hey, we're plugging that into our marketing system. And so we actually, for a while, trust hard, tried to partner up with a bunch of those other marketing, you know, software's platforms. And then we just pretty quickly realized that, you know, it's a little bit like, you know, if took a human brain and you sort of like, without being connected to like, I don't know, some arms, some legs, like a voice is like, just actually not that useful.

we realized it was actually pretty hard to integrate this with some of the older tech that was out there. And that got us into marketing, like you mentioned, and then this past, you know, a couple months ago, we expanded the surfaces, as we say, the Clavio supports to go from marketing to now customer service. We have our customer agent as well extend onto people's websites and their mobile apps with our customer hub products. And yeah, I mean, ultimately for the 180,000 brands that we serve.

Andrew BIalecki (04:16.118) Our goal is that if you want to treat every single one of your customers like the only one that matters and you want to be yourself as an entrepreneur or like at that business, your best sales rep, product expert, you can do that at scale and you can do all that through technology.

Nataraj (04:31.516) What was that, maybe initially like what was that year or moment when you felt like, okay, we hit some kind of product market fit here.

Andrew BIalecki (04:41.442) Yeah, so, well, I knew we were on the right track when, know, so we started out and I had this experience, before Klaviyo where I had all these little side projects. And, you know, remember building this, race search engine, like road race search engine. Cause I ran a lot of like five Ks and half marathons and things like that. And I remember, you know, I would, you know, the business there was like, was go talk to these race organizers, right? And I keep, I'd pitch, you know, this like search engine that we built.

And I remember doing that over and over over again. I was like, man, this is really repetitive. It'd be awesome if there was like software that could do this for me, but do it the same way I was doing it. The same enthusiasm, right? The same like, you know, level of depth of like what a product could do. And so anyways, we built this. I remember one of our first customer was this haberdashery. So for folks that don't know haberdashery, means they make, you know, kind of these bespoke, you know, custom tailored suits. You know, in their case, most of their clients are men, but some women as well. And,

I remember talking to that business and they're like, yeah, yeah, this personal touch really matters. And we built this kind of database engine that just knew everything about every single one of those that, that customer, that business's clientele. And I remember talking to them and saying like, you know, Hey, we're thinking about getting into marketing and messaging and not just being this database and source for analytics. What do you think? And they said, man, if you do that,

we'll triple the amount we spend with you. And I remember calling my co-founder after that and saying, hey, Ed, I think we're onto something here. Like this idea of being the brain of a business and being the way that they actually like talk to their customers. Like, you know, for this company that really, really cared about every little detail of the customer experience, you know, that's, think when we knew we onto something. I think after that, you know, I had a friend probably maybe six months after that that had,

you know, had just started his own business. Uh, and he was selling, uh, you know, quilts, um, you know, selling quilts online, doing e-commerce and he just spun up. He's like, Hey, there's this thing called Shopify. Have you ever heard of it? And I said, no, but I looked at it I was like, Oh, I was really impressed by, know, their APIs and the product itself. said, Hey, you know, I think we could probably integrate with this both to like kind of, you know, be this, you know, like pull in all this information into our kind of brain that we built. Um,

Andrew BIalecki (07:06.158) And then also to like, you know, integrate, you know, onto their website, and help helps off them. And anyways, I remember he's like, Yeah, I would use Klaviyo if you guys could support them. And man, I remember we built that, you know, we were connecting with a lot of different SaaS products in like 2012 2013. But I remember that just took off like crazy within like, you know, six months, 12 months. And then we're like, Okay, I think we found, you know, this really good use case. And

retail and commerce and that became, you know, kind of where we started.

Nataraj (07:39.842) Were you an app on Shopify marketplace at that point of time or how did the integration look like at that point?

Andrew BIalecki (07:46.479) Yeah, it's one of the really interesting things like I you know, I you know, this this idea that platforms have kind of an app store, you know, a little bit modeled off off of Apple and I remember, you know, Toby and the Shopify team were really early into thinking about this. Like the funny part was, I don't think it was I mean, they might have called it an app store, but it was I mean, there weren't, you know, hundreds of, you know, apps or integrations in there at the time. And so one of the things actually, I we're very strong believers.

about when you're building a business is we really like this partner model of how do customers find you? And we have this thesis that, hey, look, there's really like three things that you should aim at if you want to get discovered. there three ways to think about like, know, normally like marketing and doing demand generation. And those three are like peers. like word of mouth, like you want people to talk about you. So if your product's so good that people are like, our limousine test was always like.

Hey, if people are at like a cocktail party or they're hanging out with their friends and you know, maybe they all do something similar and somebody says, Oh, hey, like what's interesting. You want to build products that are still good. People just can't help but talk about, right. So, you know, rely on like word of mouth and peers and then, you know, partners for us. You know, we work with over 3000, you know, marketing agencies and we kind of felt like they were the experts that everybody's turning to. So if they were recommending us, it great. And then the third was platforms.

And we had this thesis that when people were building businesses, everybody obviously wants to know their crew their customers are and wants to treat them well. But we could get to a lot of those businesses faster if we could partner up with some of the platforms. And so we thought about what are the platforms that people are building businesses on? And this is around 2010, 2012. We thought that a lot more of that would be software as a service. I mean, think that's also like as I think

come to be true. But at the time, there was a lot of people that were building things on their own, or they were self-hosting, using open source. And we kind of felt like more and more businesses were just going to say, no, no, I don't want to deal with any of that. Just point me to the platform. That's a winner. you know, I, we, so we talked to dozens of these platforms and Shopify was just so forward thinking about this. You know, they had this approach. They're like, look, you know, we're going to be great at this part of the entrepreneurial journey.

Andrew BIalecki (10:09.202) And, you know, we cared a lot about, you know, entrepreneurs and folks just starting out too. Because it was in my co-founder and I's, you kind of background and roots. And, you know, we said, Hey, look, we can, I think we can help solve some of like, you know, these businesses, not just, you know, you know, setting up a website, and, you know, actually transacting on the internet, doing all the hard things that resolve around payments and fraud. I think we can be that source of truth of who your customers are and then help.

you help them do better marketing, better experiences, which will in turn will drive, you know, more growth for them. You know, we should partner up. And yeah, they were just, they were awesome. And I think back then, I've talked to bunch of folks that have talked about, how to think about these like app stores and platforms. Now, it's gotten a lot more maybe competitive or little more crowded. But back then, there actually weren't a lot of folks thinking about this way. And there were very few, like Shopify that were actively trying to curate, you know, you know, developers and software companies.

to kind of help them, you know, flesh out or, you know, make their overall product better.

Nataraj (11:10.108) Yeah, one of the things that I always think about Shopify is as it's sort of like they've looked at WordPress and removed all the complexity for e-commerce.

And you basically copy what WordPress can do and used to do, but remove all the complex part of WordPress and make it more easy to do this for an entrepreneur to not think about hosting WordPress, to not deal with plugins. Plugins evolved into App Store and Shopify, but it's a much better marketplace. It's much better integration. It's such a simple idea of abstracting away things, but it became sort of like this big opportunity when you just focus on e-commerce.

In the internet stack of things, we have these broad use cases. One is if you wanted to get customers, you use Facebook or Google Ads. So they found their product market fit there. If you want to just host anything, then you have the three clouds. And I feel like there's a small space where if you want to talk to your customers consistently once you know who your customer is, then you need to have Klaviyo or Klaviyo Live product. I feel like in that map of internet tech stack, then Klaviyo has its base, soft drama.

Andrew BIalecki (12:21.55) Yeah, like our so yeah, I totally agree. You know, some product principle we have is you can learn a lot from the folks that are maybe the early adopters, the most advanced users. Yes, we have this product principle that's like, hey, first build things to make it possible so that the folks that are pushing the limits, they can at least accomplish what they want to accomplish. So I remember for us, it was interesting when we first started. Yeah, we built this database. We built all these APIs.

And we pretty quickly realized there are a whole bunch of, you know, we actually expand our reach if we would just build some of these like connectors, these data connectors, because there were so many folks are like, I'm either technical, but I don't have time to build integration into this. And so we sort of a bunch of like, libraries, Ruby, Python, Node, PHP, et cetera. Then we realized it's like, if we actually just connect to some of these platforms, it actually would take a lot of the cognitive load and a lot of the work off for folks.

And then obviously you got a whole bunch of people that just were not technical at all. And they were like, why, by the time I go find a developer and build this thing, like I'll have already given up. So this idea of like, make it possible first, like maybe with those APIs and then make it really easy. So then you add the kind of the, know, the almost like the sugar layer on top, right? That makes it fast is the thing. And then, yeah, I think any, any great product company, know, you look at, Hey, what is the abstraction or we use this, you engineering term, what's the primitive, what's this like Lego block that you're putting out in the world?

Because you're right. Because then, people when they're, you know, when they're building any project, right, whether it's a business or startup, or maybe they're, you know, they just think about like, who has the best, you know, infrastructure and kind of building blocks. And, you know, there's really powerful thing when you start to become almost like a default, right, folks, you know, one the the crazier things that's happened at Klaviyo as we've grown is, I used to talk to our customers, and I'd ask them, you know, you hey, why did you pick us? And they'd had these very technical reasons. I tried you, I baked you off against these other products.

And increasingly, it's a little jarring. They're just like, well, why would I start anywhere else? mean, know, Klaviyo is just the best. I think about that, I'm like, man, that's like not founded in any kind of logic. I mean, that's great, but it's become almost like a brand identity of what you're good at. So we have to work to keep that. And then also, think for some of the things we're building now is like we have to basically, we have to become the building block or the primitive for those other use cases.

Nataraj (14:45.138) One of the headlines that really struck me when Cliveo went public was this, how capital efficient you were. I think you raised around $400 plus million, but you only spent like a fraction of it by the time you went public. Talk a little bit about your fundraising journey and the approach to being capital efficient, which seems to be a little bit of a rare thing in startups these days.

Andrew BIalecki (15:12.494) Yeah, it's interesting people talk now about this. Hey, you know, when were the first company with maybe, you know, I don't know, 10 folks, even a single folk, you know, achieve, I don't know, a million dollars, right, 10 million or 100 million in revenue. That was kind of our method, our thinking from the start. So my belief, my advice to founders is, especially if you're technical is you actually don't need as much capital as you think. mean, it depends a little bit on, you know, what the, you know, what

what industry or problem you're after. But in general, think people overweight, you know, how much fundraising and sort of the boost that, you know, gives you and really just boils down to like, you just got to build and you just got to, you know, you got to go nail it for some customers. And the interesting part is, once you do that, I found a lot of the best investors, the best VCs, they're sort of following, you know, customers and markets anyways. I mean, a lot of our investors over the years have told me they're like, the reason they know about Klaviyo is because they were just going in

They looked at what they felt were large market opportunities. They then went and go talk to those customers. They did their own primary research and they said like, what products do you love? And that's how, you know, we get on their radar. And I always thought that was like a wonderful way to go about it is because you're really building it up from first principles. So the trick for us was, you know, when we started, said, we just want to aim at markets that are really big. And, you know, this idea of like, hey, we can deliver, you know, if you're a business, it doesn't matter whether you have

thousand or a hundred million and consumers we can deliver every part of the customer experience You know as if you know was the founder or you know the you know the best product person the best salesperson talking to them That we just felt like you know not only we thought of that as like the CRM market But that we just felt like is a massive opportunity And then you know when we when we even looked at that and this actually was a big thing of sorts of confusion for a lot of folks that you know not our customers But really like maybe some investors

which people said like, has the CRM been done? mean, isn't that, know, what HubSpot or Salesforce and you know, what we said is like, no, no, no, we were gonna do it in a very automated way. So we're gonna aim at businesses that frankly need to use software to connect with their customers. That's why we talked about building a B2C CRM. And we said, look, like we think in the future, more and more of the customer experience is going to be more autonomous. Like software is gonna decide what experience to deliver to which person.

Andrew BIalecki (17:33.231) It's not going to be like you're setting up a human to have a conversation. So that's what we're going to go after. And we think this is going to be a monster category because if you look at like just world GDP, two thirds of it is consumers interacting with, you know, small and large businesses. And yet they sort of don't have the tools to do it. So we're to build the data infrastructure where we started and then we're going to build the marketing infrastructure. And then, hey, by the way, there's these adjacencies, you know, we're going to get into customer service and our customer agent product. And we're just going to we're going to build the entire stack.

that allows these businesses to operate and because they need software to do this, to drive customer engagement and therefore revenue, we're gonna become very sticky and a must have. when we showed that vision, along with some of the early traction we had, I think that frankly made our experience with fundraising was actually quite pleasant, right? I we mostly got to spend time talking to folks that we felt like would really add value to our thinking, would level us up.

And yeah, so it's my advice to any founders is always I think you should try to delay fundraising as much as you can You know, not everybody can do that But in general, you know two things two rules we had a clear view is we would never celebrate Fundraising milestones like we looked at those like those are non events. We'd celebrate a lot of customer milestones Maybe revenue milestones, but we never celebrate fundraising milestones. And the other thing is we never celebrated You know how many people worked at Clevio? I talked to some other founders and they celebrate. Yeah, I've got a hundred people

or I've got a thousand people. And we always looked at that as like, well, that's, I that's cool, but wouldn't it be better if you were like a smaller team? Like, wouldn't it feel more intimate? You know, people bet, you know, people like the communication tax would be less. So I think that's something we've tried to, even as now we're a public company, we really try to stick to is, you know, we don't celebrate fundraising milestones, which to me is like, you don't stare at the stock price. And then we also, we just don't look at like, hey, how many people work at Klaviyo? We think about like, hey, smaller is better. And that's why we really believe in like, you know, the power of small teams.

Nataraj (19:33.299) I mean, you said you wanted to delay fundraising, which is a little bit counterintuitive to the idea of going public, because most companies go public to fundraise. Obviously, that might not be the reason for you. So talk to me about that position, because I used to work for this company called Epic Systems. don't know if you know about it. It's the health care software company.

Andrew BIalecki (19:51.085) Yeah.

It's an incredible company, incredible business.

Nataraj (19:56.243) Incredible business, pretty much 70 to 80 % of US individuals, healthcare records are managed by that company. The founder, Judy Faulkner, still runs the company. It started in 1979.

So incredible company, but she had this thesis of like never to go public. And I think she also ensured that the company will not go public posthumously also. So she had a very strong point of view. And now we are seeing like Stripe doesn't go public. There's some other crazy, good companies that don't go public for different reasons. What was your thought process when the time came to decide whether you should go public or not?

Andrew BIalecki (20:35.064) Yeah, did. I had studied all of the, we felt like all time great companies, software companies, technology companies. And the vast majority of them, but not 100%, because your point, Epic was one of the examples there. There's a couple of others. We look at these very large private technology companies, but most of them went public. And I kind of looked at that as like, OK, it's a bit of a rite of passage.

There's some other things like it gives liquidity to investors and obviously like folks at Clivia that have worked very hard for many years. This is before the private markets were maybe a little more liquid. So I also think that public markets give you a chance to just tell your story more directly. I guess there's nothing stopping a private company from doing press releases or sharing their numbers, which I know a bunch do. But I also felt it's like there's a bit of trust you can build up when you say like, here, just look at our books. You can tell how healthy our business is.

there's going be some accountability every quarter, every year. So it's interesting, like now that the private markets have gotten, I think it's they're almost operating as these like pseudo public markets. I think, you know, I don't think it changes our calculus, but I think it sort of it makes things things are almost merging a little bit. But for us, the most important thing was when we knew I felt like we would go public at some point for all those reasons.

We just didn't want to change the way we were going to operate. And that meant, you know, having a real strong, like long-term point of view, being very like product and customer focused. And, you know, I don't know, frankly, we just felt like because of the strength of our culture, like there just wasn't a lot of risk, you know, in doing that. You know, had a saying back in 2020 when we were thinking about like, Hey, when we would go public and, you know, would we and all this stuff. We said like, look, there's some all time great companies. Like if you'll get like Microsoft or Apple and you know, these companies is like,

I think everybody knows them as these like, you know, kind of standard candles for tech. And yet if you said like, hey, what year do they go public? Everybody probably has a vague sense of when it was, but nobody knows the date and nobody really cares because a lot of their best work has come since. So clearly they have like, you know, cultures and a way of working that just has stood the test of time.

Nataraj (22:53.2) I mean, yeah, from perspective of retail investor, think it makes more sense for more companies to go early public. Like, you could invest in Amazon so early, but you couldn't do that now with OpenAI or like Anthropic, right? Which probably will never go public.

Can you talk a little bit about, I think we are now in December, which is like, we just passed probably the biggest holiday cycle. We often call it as BFCM, Black Friday, Cyber Monday. Can you talk a little bit about how have things been this BMCM for Clavio and in general, any trends that you notice in e-commerce or in generally broad economy?

Andrew BIalecki (23:34.169) Yeah, yeah. So yeah, we just had a great, you know, so for folks that don't know in, you know, retail and e commerce, you know, the reason they call it, you know, Black Friday is there's a whole bunch of retail businesses that literally go into the black like this is the time of year where they turn profitable for the full year, because sales really ramp up, you know, in October, in November, and then, you know, a lot of businesses then, you know, run promotions leading up into the holidays at the end of the year.

So it's like a critical, it's a critical moment. think there's something like, it's like 10 or 10 or 20 % of our customers will do, you know, something like 30 or 40 % of their total annual revenue, like in these like, you know, eight weeks. So it's a crazy high percentage. So it's very, you know, it's a little bit like a, we talk about it's like a little bit like a Super Bowl moment for our customers and therefore, you know, for us. So we had a great holiday season, you know, our database, I think, you know, our stats were

You know at peak we were ingesting more than you know 10 billion These like consumer touch points data points into our database in real time You know, think I was over the course like that was our peak like, you know on a given day And then over the whole holiday weekend, you know, we sent more than 20 billion messages or power more than 20 billion experiences And so that you know ranges from our marketing products. It's now our customer customer service our customer agent products. That's you know on fuels websites

or that you can email or text with that will give you advice or help you like, you know, troubleshoot problems. So we're really proud of that. And then most importantly, like we've always looked at, know, Klaviyo is like a revenue engine for both small businesses and you know, some of the world's leading brands and you know, across five day stretch, we generated, I think the number was $3.8 billion in sales that were like directly tied back to experiences that Klaviyo powered, you know, that either AI or our users defined.

And that's like an incredible number. think it's about, you 5 % of like all retail sales, you know, that happened, you know, in the US, you know, globally is because we kind of cut that. That's a global number, but we cut that down. So you think about it, like, I mean, that's a huge chunk of, you know, of revenue that your business otherwise just wouldn't have. And one of the things we love is that it's like, it's highly profitable revenue because it's from businesses and their relationships with consumers. So it was a really great,

Andrew BIalecki (25:58.607) you know, weak, you know, brands did a great job, our partners, our customers, our team did a great job. I think the most exciting trend, though, to your question is really the rise of the retail brand as not just, you know, place to go buy products, or in some cases to buy services, but really those companies evolving into value add because of the expertise and services that they offer around their products.

So I'll give you like a couple of examples of things we've seen in the last couple of weeks. So we work with a lot of apparel brands and there's a swimwear brand that I was talking with and they're saying like, yeah, now's the time of year, this and the summer are of, they're two busiest times of year. And they were telling me how they were using our customer agent. So this chat agent that you can put on your website and customers were conversing with it, not just about which swimsuits to buy.

but literally having conversations about how to plan a great vacation. Because there were a of folks that were buying for the holidays and maybe, you know, if you're here in the U S for like, you know, a kind of spring break trip they were going to take. And they were asking questions like, okay, great. Well, I'm into these patterns, you know, what swimwear do you think I should buy? And by the way, I'm going away for a week. Should I buy one swimsuit or two swimsuits? And by the way, you know, here's a couple of locations I'm going, you know,

Do you have a perspective on the style of kind of, whether I'm going to Florida, the Caribbean or Hawaii? I mean, it's crazy, but like people were having real conversations, you know, with these agents that these businesses are providing. And all of a sudden now the experience or the value add these brands are, you know, are offering is not just, Hey, we make really high quality products. They look great. But it's literally not just, Hey, what's a great swimsuit to buy? It's how do I have a great vacation? So we're seeing this like,

everywhere. There's a you know, education company, you know, that I've known for a long time that used to use Klaviyo, you know, kind of caters to helping kids learn like reading and math. And one of the things they built in Klaviyo was this like, awesome sequence of like, almost like worksheets and coloring pages for kids and say maybe between the ages of like five and 10. And I remember they build these in Klaviyo and they generate all this content, you know, by hands.

Andrew BIalecki (28:20.908) And then they'd sort of set it up in Klaviyo to automatically go to folks, you know, as they became customers or consume products or maybe when they subscribed. And now they're doing the same thing, but it's all personalized based on context they have inside of Klaviyo. And then using our marketing agent, I mean, literally we'll send different content to each person based on, you know, I don't know if you're like my kids, like, Hey, are you into Paw Patrol? Are you into like, you know, the K-pop demon hunters that are really popular right now?

And all of a sudden, these like themed and even like at the level of what that individual like student wants. I mean, they've transformed themselves from being just purely like, hey, we'll help provide workbooks for your kids to practice into literally we're acting almost as a pseudo teacher and like lesson planner. So I think it's a really interesting thing is the first year that we've seen a lot more of that show up. And obviously one of the cool things for us is we think increasingly those experiences

rather than our customers, whether they're marketers, or they're product folks, or they're folks in operations, rather than them having to point and click through software to define those experiences, increasingly folks are saying, hey, look, I have a goal, right? My goal is to delight my customers, to have them be engaged. I know that translates with how much time they spend with my products or maybe how much they buy. They're giving that goal to AI and just saying, hey, why don't you come up with the ideas?

And then letting AI, with some of the brand controls and checks that we have, just test that stuff out. And it's just, I think it's a total mindset shift, right? In terms of, one, how businesses work, in terms of how much autonomy or agency they're giving to these algorithms. But then also this idea that like, it's not enough to just be products on a shelf. You need to be value add in and around the products and services that you offer.

Nataraj (30:16.218) It almost, that idea can even extend that even the idea is also coming from AI after a point of file based on years of data that a brand has of Gravia. You can just say, at this point of the year, this might work for you. And you basically come up with an idea and suggest the customer that.

for this idea, these campaigns will work. So you can almost say that the customer doesn't actually have to come up with ideas and Cravia AI will come up with those ideas. I think that's a good segue to talk about more about what you guys are doing with Cravia AI. I think you also launched a marketing agent. Can you talk a little bit about that?

Andrew BIalecki (30:51.106) Yeah, sure. Yeah, so it's a great example. So when we think about artificial intelligence and machine learning inside of Klaviyo, we have these two agents that we've launched. So we launched back in September. One is our marketing agent, which will help you do marketing. But I think critically for us, we think about our marketing agent is so good that for some use cases, it can just do the entire process of marketing for you. So I'll talk about that a bit in a second.

And our second is our customer agent, is, you know, hey, it uses all of the context you have about each of your individual customers. So that, you know, when you are not you or I land on the website and we know it's you, that customer agent is ready and primed to answer questions that you have based on your own personal context. You know, in retail, it might be things like product recommendations. Hey, you've seen my past purchase history. What would you recommend for me? Or it might be things as you know, helping with some things operationally.

Hey, I ordered something, but I actually got the address wrong. you know, it's actually, it's a gift for somebody. So I need to change it from my address to somebody else's and our agent can handle those types of queries. So for our marketing agent, you know, one thing, if you're building agents, I think that we really lean on is this idea of mapping out the actual process of, whatever task you're doing. So one thing that I think we have sort of a unique advantage on is we've been studying how people think about marketing and marketing is this

It's this great mix of like, it's brand, it's feel, it's also just creativity and the kinds of campaigns you run, but it's also just pure math and performance. Hey, what do people actually, you know, how do I know whether somebody likes this? I can run qualitative surveys, but also just how do I actually see what people are, you know, kind of voting with their feet or their clicks? So what our marketing agent does is it breaks down the process of marketing into a couple of steps. First, it'll go through all the information that's stored in Klaviyo and even some external sources and

Figure out trends and patterns, we call them data insights, where there's something interesting that might be happening. So I'll give you a couple of examples. It could be things like, hey, we noticed that you just added a bunch of new products, say, into inventory. We should generate a campaign around that. OK, that's probably pretty obvious. People would know that anyways. We're also good at finding things like, hey, here's a bunch of products that aren't selling through. Maybe you should try to clear that inventory because it's taking up space in your warehouse or you're paying for it.

Andrew BIalecki (33:14.494) And hey, the seasons are shifting, and it's time to move that inventory along. Or we can find things like, hey, here's customers that they loved the starter version of your product. For example, take our educational company. Hey, you sold them a workbook about math that's maybe geared towards first or second graders. But hey, we actually found an opportunity where those kids have probably grown up. It might be time to offer them some of your content that's geared towards older kids.

So it'll go through and find all of those insights. We then convert those into what we call marketing briefs. So this idea then of like, how would you actually build a campaign around this? And it almost like writes, it's a little bit like a product requirements doc. It kind of writes down what it thinks it should do. And then we have this really great set of algorithms and agents that will then do the actual creative generation. So you can think about we're used to AI these days. It can generate copy, can generate imagery.

But we do way more than that. We help define larger structure, like, hey, what does it mean to design, say, a whole email or email plus a text message plus a WhatsApp message. And WhatsApp, you can have all these cool carousels. So it's like it can do some of these higher order objects. It'll also go do incremental research, right, where it needs to create content so it's smart about what it's offering. And then finally, we actually use AI to play a little bit, be a little adversarial and say, OK, now that we've created all this content,

hey, rank it and also make sure that it passes all these brand guidelines that we have. And some of these are basic checks around like kind of content safety and toxicity, but also just correctness. And then like we can actually, we can actually pull together a kind of brand, a brand guide for every business just based on things like their website, internal documentation they have. And you can imagine all of that's happening. And literally what we present to our users is, okay, here's all the ideas we came up with.

Which of these, you like you get, now get the chance to like go approve them. You know, are these ready to go? And literally in a couple of clicks, have customers are saying, wow, you create a full launch, you know, campaign for me. Like, Hey, I might make a couple of tweaks, but then yes, go schedule that and let's see what happens. And then, you know, using reinforcement learning, we go watch, you know, Hey, which campaigns, which of these, you know, marketing experiences pervert, convert best. And then can feed that back into our algorithms. So I think this is kind of

Andrew BIalecki (35:37.571) you this is really like the future and it solves one of the number one pain points I hear from our customers. Hey, you know, I feel like, you know, Klaviyo is really great software, but I just don't know what to do with it. Right. I have some ideas, but I wish I had more ideas. And now I think with, you know, LLMs, we're able to basically automate the entire process. And then our customers can kind of decide, I think today they want, you know, a high degree of agency, a lot of checking of the content we're creating.

But what we're finding is as the quality gets better and better and as they build trust, we're starting to see them almost like auto-approve. We have some campaigns we see users, they'll look at for maybe 10, 15 seconds, and then they'll send out to hundreds of thousands of people. So there's a lot of trust being built up quickly. And I think this is, frankly, it's not just the future of our category in this kind of consumer CRM space, but think all software, a lot more software is just gonna be able to run itself. And then you can choose how much you wanna be in the flow.

Nataraj (36:34.098) Is there any particular metrics that are interesting or that tell whether all these initiatives are successful? in the case, in the example that you gave, it could be just how much time it took for creating new campaigns. Do you guys track specifically some measurements, like how successful have been these, either the agents or other AI products that we're launching are successful for customers? What is the way you think about that?

Andrew BIalecki (37:01.272) Yeah, I'll give you two examples. So, first is on the quality. One thing we're just obsessive about is, and we know this, is when we first built our marketing agent, we kind of said, it's okay, it's like good marketing intern. And that's, no disrespect to all the very highly motivated folks that are breaking into marketing, but we just felt like, okay, it's not that knowledgeable about best practices. We're still building that in. And what we feel like is, hey, it's going to go from that of intern level to, you

Maybe like a senior marketer to like a really advanced marketer and eventually if you think about you know Some of the media types that we're working with but you know text messaging email mobile Social I mean not all these are only like 20 years old We think it's soon gonna be able to beat you somebody who's been spent. know the last two decades Doing nothing even this so in the way we measure quality there is it really does you know we do one is evaluate like hey how much do the folks that were effectively pitching our ideas to so our customers do they like them how much are they adopting them

But then how are they performing? And what's really cool is we're seeing, you know, in a bunch of cases, our content that we generate is outperforming in terms of engagements. You can think about like click rate, but also just literally then in terms of downstream engagement, in terms of like purchase rate or conversion to customer. So that quality metric is something that we really obsess about. And then we want to show to folks so they can feel confident that like they're not losing anything. They're actually gaining things by sharing their ideas with us or adopting some of the ideas that our AI is producing.

And then the other part is straight up, how much do people are starting to use our marketing agent and our customer agent as part of their everyday workflow? So the analogy we make is when mobile phones, starting with the iPhone, really got real web browsers, Safari for the iPhone came out. There was really interesting charts. They said, OK, let's look at internet usage and how much of it is coming from desktop, or we think traditional desktop versus mobile.

And if you look at those graphs, something really interesting is it wasn't, you know, it wasn't that desktop usage, you know, went away. I it's probably declined a little bit since, but it really that was that mobile usage exploded. internet usage went up. So the other thing that we measure is, hey, when you look at the number of ideas, marketing, you know, campaigns that people are running, or the case of our customer agent, hey, how many conversations is this business having with its customers? We track both of those and how much of those are say,

Andrew BIalecki (39:27.596) human generated, human powered or human initiated versus how many are AI initiated. And what we're finding is actually the exact same thing. Like it actually turns out that when you introduce AI, it's increasing the overall amount of marketing or the number of customer service and number of conversations that businesses are having with consumers. And it actually has this interesting thing where because you can kind of, you can do this hybrid model where you use, know, AI to either make your idea better.

Or you use AI as maybe the starting point and then you know and then the actual human, know can kind of go edit from there, right? We're seeing that both with marketing and with customer service customer service We these great things where I can do the first bit of like kind of collecting all the information that's maybe relevant to helping solve somebody's problem and rather than that kind of back and forth of like, okay Can you remind me your email address and can remind me what you're talking about like all that just gets automated? We're finding that's driving up the quantity

of both marketing experiences and customer service experiences. And that's another one that we're really bullish on is that over time, we do think that, know, not like it replaces, displaces what, you know, human thinking is doing. It's actually just, it's literally just growing the overall pie and then the amount of, you know, the amount of engagement that, you know, brands get with their end consumers.

Nataraj (40:46.066) It's almost, I think there's a term for this, it's called Javan's paradox, when you make things cheaper, the overall usage goes up. And I think you're basically making it less costly in terms of how much time I spent to create.

campaign so I can quickly create cameras and just verify, okay, this campaign looks I want to experiment and just giving more time to the customer and that's how they're using it more. The other question I had about AI is did anything in terms of how you organize the company or what types of roles you're hiring or how you're managing people change because of AI and because pretty much every startup founder says we are either hiring less people or

We are doing things differently or we are hiring for a different set of people. We want more generalists, we want specialists. How has CREMEA changed? Or it hasn't changed, basically for AI.

Andrew BIalecki (41:44.271) Yeah, so it's really interesting. it's fine because we so we bootstrapped our business. Like we're talking about when we started, which meant we didn't raise any capital until we were already profitable. like, again, I would just you know, not every business can do that. But I'd say like, oh, man, you should really try to run lean. And why do that? It's not just a financial construct. We found embedded in our DNA a couple of really great attributes. One, everybody had to be a bit of a generalist. You had to be broad. you know, I

There's a saying that you kind of look for when you're hiring people, look for people that have this kind of T shapes, you know, kind of skill set. Hey, they're really good at something. They're kind of okay at others. We tell people that we like people that are like, you know, sort of have a very wide T, right? They're almost like a fat T shape. They're like, they're really good at a bunch of things, right? They're not sort of a superstar at one and then okay at others. They're like, they're actually pretty good or great at a couple of skills. And that intersection makes you like even more dangerous. So when it comes to, you know, let's say engineering, you know,

when I talk to engineers, say, look, you know, I want, it's important to me that you care about design, that you think about the products you're building, by the way, that's way more interesting way to live life. Like if you just want to build really scalable systems, like that's sort of like not interesting all the way, right? So our, you know, all of our engineering team, like, Hey, we're just into, you know, talking to customers and nerding out of that. And actually I think you build better products that way. So generalists matter. And then two is, you know, the constraint we had for bootstrapping was, you know, a job was expensive. Like I remember in the early days,

My workflow was I'd wake up in the morning, I'd spend anywhere from 30 minutes to a few hours in our help desk, just our inbox, like answering customer questions. And I wasn't allowed to start building, as I thought, making real progress forward work, until we had finished helping all of our customers. And that really forced you to focus on like, okay, what are the real pain points that…

users are having and how can I automate? In a lot of cases, people were just looking for better documentation. So it got us to write better documentation. So just force this idea of like, how do you automate yourself? How do you automate yourself? How can you give yourself more scale, more leverage so you can not spend your time on tasks that are very repetitive and get to like, know, higher order bits? I think AI is just another, you know, incantation of this. So well, those two attributes have stayed very true to how we work and felt very natural with AI. And we do a little bit to like, you know, we oftentimes will force constraint.

Andrew BIalecki (44:04.802) you know, on teams and projects will say, look, we're going to ask the small team to do something really big. And it feels hard. And think a lot of other companies may say, well, that's impossible. But for us, it's like, we just have this track record of we've seen it work. Right. So we're like more willing to do that. You know, the one thing that with AI, I think we had to do is we sort of had to force ourselves, say like, OK, we've been pretty successful as this, you know, software as a service, Internet, you technology company. But

AI is just going to change a lot of workflows. if you sort of feel like, I I'll take myself, I just got so used to it. like, I mean, I love building, love programming. But if I like writing the code so much that I'm sort of not willing to try my hand at letting an AI do that, that's probably a mistake. And, you know, and hey, we really need to lean into that. one of the things that we'd obviously say, hey, look, everybody just needs to be, you need to sort of force using LLMs into your life. And

This is actually, it's kind of annoying because, you know, as we've seen, like, LMs are not great at everything. You know, they make mistakes, you know, they used to hallucinate a lot. They still do in some cases. And I think there's just, you know, I remember in the early days of the, you know, the internet when I was building websites and web apps, like there was a lot of things where it was hard to do things on the internet, you know, building a highly responsive, you know, website, whether it's cause you're using Ajax, right, or using modern CSS, like it was hard, but these were just men, but we had to figure out some of the frameworks and build some of the tooling.

I think AI is the same way. So we sort of forced everybody to say like, hey, look, if you're just not naturally the kind of person that wants to go after, you know, building LLMs into your life, then like, hey, that's probably not a good fit. And the analogy we use is like, imagine that, you know, it's the mid 90s, and you're Amazon and you're hiring people to join Amazon. And imagine they said like, yeah, I don't really use the internet. I don't really know what a browser is. You're like, probably the wrong company, right? And

I'd say even if you're just like, well, somebody showed up to an Amazon interview and said, well, yeah, I use Amazon because I heard about it on the news. You'd probably say, okay, but like a lot of people use Amazon. So the analogy for us is like, well, a of people use chat GPT. Like that's, actually not that novel. What we want is the nineties example would be somebody that built their own website, you know, even maybe a personal website. That was like hard to do back then. I mean, now it's like, Hey, are you building LLM applications? You know, and by the way, you don't even need to be an engineer to do that anymore. You can just go sign up for

Andrew BIalecki (46:25.038) one of the many, many, many tools out there that allow you to like basically, you know, WYSIWYG, you know, vibe code something. If you're the kind of person that does that naturally, awesome, can't wait to work with you. We're have a great time. If you're not the kind of person that does that, then like, hey, you know, maybe you're not really into, you know, you're not really ready to give this a go. So I think that that's been important, is having density of like this kind of curiosity and, you know, thinking, you know, AI first.

Nataraj (46:50.172) think the point you made about automating yourself, I think that really hit the nail for me because that's sort of like the most obsession I have right now is…

How do I automate things? do repeatedly. And AI is sort of like open up and you sort of 10x the amount of automation that you could do as an individual. And that's really amazing. Like how many things you can, it's literally that code of like, give me a lever and give me a truck to place so I can move the earth kind of scenario right now. Because like you can do so many things.

just one person because of so many tools or you can just, if you don't know a little bit of coding, then you become so much more powerful than you ever were. regarding the, you know, going general this way, one of the things I found out is like, create a really courageous people to have their own sidekicks and hiding people who are like, you know, either influencers or, you know, have their own small business. Tell me about the like logic of that.

Andrew BIalecki (47:51.407) Yeah, well, I think, so this founder mentality or entrepreneur mentality, first of all, I don't think you have to literally have gone and started your own business to earn that credibility, to earn that title. It's really a mindset. It's certainly a great, I mean, if you do strike it out on your own, and by the way, I think a lot of folks, if you're gonna start something, my big recommendation is do it nights and weekends, because then it's hard, right? It's like extra time.

tired, you maybe have other things you want to do in your life. And like if you really want it, like, and you put in the hours then then you sort of know that you're ready to do it full time kind of thing.

Nataraj (48:28.09) Michael Lewis, the author of all the famous finance books said the same thing about becoming an author. You should write it on vegans and see if it actually works. But good.

Andrew BIalecki (48:40.974) Totally. mean, before we started Klaviyo, mean, one the things I did to kind of steel myself was I was like, okay, I'm going to try building these, you know, little side projects and like, look, I mean, it's no sweat, right? I'm still, you know, I'm still getting paid at my current job. But if I can't find, you know, the 10, 20 hours a week that I'm and I'm not so excited about it, then like, that's probably not a good sign because it's, it's only gonna be harder, right? When you know, that's the only thing. So it is we look for folks that have these entrepreneurial like tendencies.

And anyways, it's great signal if you know, you've actually you I mean, oftentimes you will show me a side project, you know, hey, here's a GitHub, something I'm working on. Hey, here's this thing I'm doing in my community that I didn't necessarily have to take on. I think entrepreneurial density is really, really important. You know, it offers tend to take high degree of ownership. You know, there's there is nobody else to help them because it's like their thing, right? If they don't solve it, nobody's going to come solve it for them. Yeah. And so we've extended that like so one, mean, we try to make

Klaviyo a great place for entrepreneurs, either aspiring or folks that have been entrepreneurs in the past. I was talking to the head of our customer agent products, the guy that leads all of product development for that, who is himself a multiple time successful entrepreneur. And he said, he's like, yeah, one of his favorite kind of recruiting pitches is like, hey, Klaviyo is a great place for aspiring or former entrepreneurs that just want to do the same thing.

but hey, they just want to do it inside of another company, right? And so we've then tried to architect Klaviyo where I think at its best, I was a physics and astronomy major, so I've always liked this analogy of the best companies as they scale, they feel like a constellation of smaller companies, which means a lot of autonomy for teams, right? So it's a big responsibility, but also it means you give them room to, maybe we're all bound by the same kind of gravitational field, so we're all generally going in the same direction, but it's okay, some stars maybe moving in slightly different,

orbits and that's okay. So we give teams like a lot of autonomy. And then like, you know, we just say like, Hey, like, let's agree on like, sort of what the goal is and where we're aiming. But after that, like you be the entrepreneur. And then, yeah, I think this surprises some folks, but like, you know, I've had over the course of Klaviyo, probably, you know, maybe a dozen, two dozen folks, you know, come to my desk, right? And say, Hey, Andrew, I have some bad news. Like I'm quitting, because I'm just gonna, I really want to go try, you know, building this product, starting this company.

Andrew BIalecki (51:06.19) And a lot of those are software technology companies, but like we had a couple years ago, we had somebody that like was early at Klaviyo and he said, hey, I really want to try my hand at running a cranberry bog. I said, wow, that's crazy. You want to run a farm and like, you know, specifically for like cranberries, that's, that's like awesome. He's like, yes, I just have this passion. I really want to try it. Right. And of course he was like super entrepreneurial. Right.

And every time I think, you some folks are like, man, I can't believe it. You know, why would you do that? But I have a hard time feeling that way because I I literally worked at other companies and then eventually left those to go start something. Right. And I think frankly, in the world, we just need more entrepreneurs. You know, we talk about creators, people that are just creating products, services, content. We literally need more people that are right. You're Michael Lewis are willing to are willing to put in that hard work. Like we need more people that want to take those shots.

And so we think about Klaviyo as a culture, as like, okay, we want to help develop those future entrepreneurs, founders, leaders that think very independently. And then obviously our product mission is, yeah, literally our mission statement is to empower more creators to own their destiny. We literally want to give more businesses the tools to be independently successful. So this idea of like entrepreneurs, it's just like everywhere throughout our culture.

Nataraj (52:23.826) I think that's a good note to end the conversation. think, know, Klaviyo has been, you know, anyone I ask about Klaviyo when I was researching this, you know, everyone has good things to say and I think your NPS scores must be really high. But, know, Andrew, thanks for coming on the show and sharing audio, all the interesting things about Klaviyo and I'm looking forward to see, you know, where Klaviyo goes next.

Andrew BIalecki (52:49.176) Cheers, thanks for having me.

Full Transcript

Nataraj (00:01.87) Hello everyone, welcome to Startup Project, where we deep dive into the minds of innovators and entrepreneurs that are shaping the future of tech. My guest today is Ilan Reshaf, the co-founder and chief product officer of Gong.io. Before co-founding Gong in 2014, he was already a seasoned entrepreneur, having co-founded and led WebCollage, a successful SaaS platform, which was acquired in 2013.

Gong leverages advanced AI to analyze customer interactions and sales conversations and enables teams to boost their productivity, deliver revenue predictably, and drive efficient growth. Under Elon's leadership, Gong has evolved from initial conversational intelligence offering into a sophisticated, derivative AI operating system. It reunifies customer insights through proprietary Gong revenue grasp, diverse actionable intelligence, and automates critical workloads.

have over 5,000 customers, including prominent names like DocSign, PayPal. Gong's impact is unbind and deniable. It helps companies achieve outcomes like 57 % higher win rates and saves thousands of operational hours. Today, we'll explore the genesis of Gong, how they achieve product market feed, and how AI changed their business and product, and a lot more other interesting things. With that, Elan, welcome to the show.

Eilon Reshef (01:26.306) Thanks for having me.

Nataraj (01:28.155) So I think the first question I wanted to ask was what was that initial problem that sort of like caught you and your co-founders attention that led to GUNS?

Eilon Reshef (01:40.768) It's a good question. And actually Gong is one of those boring companies where not much has changed in terms of the overall vision from when we started and today, despite all of the revolution that happened within the technology world, obviously LLMs and whatnot. When we started, that was about 10 years ago, 2015. And at the time, we're looking at the revenue space.

And what we noticed was people were, it was treated like an art, right? Sales is art. And we felt strongly that if you can introduce a time, we didn't even call it AI, was data science, data-driven workflows, whatever you want to call it. You can make things much more efficient. again, today everybody's talking about productivity and AI, same idea. And then we realized that in order for quote unquote AI to make sense and help.

First of all, you have to have like quality data, which hasn't changed. AI is only as good as the data it gets. So what we started as a company was let's capture the core information, which some people might think is the CRM, but in reality, it's actually the conversations that people have with customers. So it could be sales, sales could be post sales, be sales engineers, SDR, whatever. And the idea is if AI has access to those conversations, then we can start making like really, really good decisions, recommendations, and actually carry out actions for you. So that's exactly what we started.

At the time, we kind of hooked up to WebEx. That was our first video conference system. Zoom was barely starting. And then later we did email and text messages and other data sources, and of course, connecting to CRMs. But the core genesis was let's bring information, put it in some sort of, and we call it a revenue graph, some sort of a graph system, and then apply logic to it to help people be more productive and leaders get more intelligence.

Nataraj (03:21.805) think in some sense a lot of these ideas are now a little bit more common.

But I think back then it was not that common. Getting all the call data and transcribing it was not really native to any of these, if you're using Zoom or anything. Transcription was not native. recording calls was native, but we didn't transcribe calls. Even if you use Teams or Google Stack, transcription and getting that data and analyzing that data was not really common. Now there's a lot more competition in that space. What was the tech stack?

for when you are using, you're trying to solve this problem.

Eilon Reshef (04:01.282) It's a very good point. Even the idea of recording at the time, people did not want to invest in Gong back in 2015. I mean, obviously we did raise money, but most of the majority of VC firms did not want to invest in Gong because their hypothesis was people will not want to get recorded. Which was of course right now we're entering 2026 and obviously it almost feels like ridiculous at this stage, but it wasn't obvious at the time. And what we had to do was actually invent or develop the core recording technology because most providers

I don't could not record even calls even if they did it was some clumsy process where maybe a user had to manually click record which of course wouldn't So we started with technology stack that's basically developing a bot that joins the calls now It's so common that yeah, sometimes if you're meeting like four people and an 18 pots or whatnot But at the time we still have a patent on what's joining call. So we developed this back in 2015 And it's joining the calls. It's sort of the automation around starting recording capturing the screen whatnot, of course keep bringing it to the back end of

Of course, it was a robust kind cloud first, you know, kind of system to begin with, 2015 modern stack. And then in the beginning, we used the third party transcription engine, which really, really sucked at the time. It wasn't because of the system. It was just because technology in 2015 was like, I think something like 30 % word error rate, which means three out of 10 words are actually wrong. So you can even read the transcript. The first versions of GoG, we essentially hit the transcript beyond like four clicks so people wouldn't find it. You could still search it and do statistics and like high level topic detection.

but like reading it was really really hard and then very quickly moved to a homegrown system which was better and of course nowadays we still use a homegrown system but it's much more much easier to just kind of I don't know take whisper or any of the major providers transcription off the shelf and then kind of use it and get I guess pretty good results.

Nataraj (05:50.319) Who were the early customer adopters? you mentioned like, transcription would be not that great, but like who, which type of customers really leaned in on this?

Eilon Reshef (06:01.778) Yeah, so I'm a big believer in sort of the crossing the chasm, me and my founder both are big believers in crossing the chasm model, which means you want to start with a very small niche. And so when we started, we basically said, who's adapting technology fast? It's like technology companies, good. Which technology companies are more likely to use video conferencing? Like guess what software companies, because they don't want to travel to.

you know, destination because software it's like easier to sell software online and physical goods. then, so we pretty much said, Hey, and enterprise we couldn't sell because we didn't have like a way to set to enterprise. was just like, we didn't have security. didn't have scalability. So we said, let's start with software as a service companies in the United States of America, North America, selling in English over video conferencing, midsize companies selling midsize ticket items. Because if you're just selling, I don't know, $5, you probably the conversation is not as

important to you. Maybe even the wholesale cycle is more like a B2C type of thing. And then if you're selling, I don't know, if you're buying and selling airplanes, my guess is you're probably not going to have, like you're to have much more a relationship selling and in-person selling, which at the time we couldn't support. So the idea was like, focus there.

And then afterward we said, hey, it's not only video conferencing, it's phone conferencing. And then you pick an email and then you can start doing more relationship understanding. And then you can have like, now we have an in-person recording and all sorts of other things. But the idea at the time was like, focus, focus, focus. There's probably 10,000 companies in the world that focus on this category.

But look, to get from seed round to A round or to B round, whatever, you only need like a dozen, two dozen customers. As long as you know there's enough companies in the future, we're very satisfied. And just like starting with a very narrow customer base.

Nataraj (07:45.007) What kind of early insights was gone providing at that time, like in 2015, 16, whereas adopting, because prescription itself was not accurate, what does was actually find valuable?

Eilon Reshef (08:00.49) Yeah, so if you think about what can you do with this transcription that is not accurate, it almost lends itself to what are the killer apps for this. And one killer app was Search.

So even if you don't, like some words are missing, you search for a competitor, you still find them. So we invented this idea called a tracker, which still by the way is available in every Gong, I don't know, wannabe or even like a big, actually even Microsoft has some sort of a conversation that is this product that actually even use the word tracker inside it. And the idea was the tracker was like a saved search or a saved keyword list.

And then what you could do as an organization say, hey, I want to focus on those conversations that actually use, you know, bring up competitor X or challenge Y or technology Z or whatever. And then you can program the system for this to kind of drive many, many workflows. One is just like, I'm a product manager. want to know what's happening in the field or very common use case. I'm a sales manager. want to coach people, but I don't want to coach them on every call. want to call it just like talks about pricing. Or I want to help my team.

position against the competitor and I want to find only conversations that are about the certain competitors. So search was a phenomenal use case or narrowing down filtering this side.

you know, very big graph is sort of like super important. The other one is more of a collaboration use case, which is I'm a salesperson. I got asked the question, I need to bring in more people in the loop and I can start tagging people and just having a very convenient interface where you can bring more people to collaborate. There's a chat window, there's you know, kind of mark moment 27 and it's making it very easy to people to kind of.

Eilon Reshef (09:44.254) send as a team and that was a very use case. So these are maybe the two main ones, one more AI, one less AI maybe.

Nataraj (09:51.375) And when was like that sort of like tipping point where you thought, okay, we've achieved product market, but was it like very early on or like a couple of years into the development of the product?

Eilon Reshef (10:03.106) Yeah, that's one of our funnier stories in Gong. So this is going to be like maybe two different answers. So one is when should we have realized that we product market feed and then maybe, you know, us being a little bit slow when we actually realize it, right?

So the point when we should have realized it, we raised money in October of 2025. I brought some of the team members who worked with me in my previous life. So January, three months in, we had like a running prototype that could record many calls and transcribe them and do all of the things we just discussed. So we started giving it to customers, again, alpha, beta customers, whatever, just like SaaS companies, know, kind of our size. I mean, not like five people, but whatever, 500 people, maybe 1000 people. And then we gave it to 12 design partners and we

them, could you please give us feedback? And they started complaining and we fixed things. Of course, nothing worked in the beginning. And then at some point they stopped complaining. And then when they stopped complaining, we were like, why are you stopping complaining? And we watched obviously their behavior and they're like, we're using it. It's fine. Why are we complaining? And then I meet my co-founder basically said, what if they are not complaining and are using it? Maybe we start asking for money. Right. So we had 12 design partners. We called them and said the beta is over. There wasn't any beta, right? It was just like, we just gave them the software to try it out. Right.

And then 11 out of the 12 paid and that was like May 2016, which is six, seven months into the company, right? And 11 out of the 12 actually went ahead and paid. And we weren't cheap at the time. were like charging initially maybe 750 per individual per year. like this price now is higher, but you know, for a young startup with a dozen employees, that's not small, right? And 11 out of the 12 paid and then the 12 actually paid a year later. I mean bought a year later, the CROs changed jobs and like they couldn't buy.

And I think that's probably a point where you should stop and be like, shit, this thing is actually working. Cause like 11 out of 12 is like unreal in a way. But we're like, okay, that makes sense. Let's maybe, I don't know, why isn't the 12th buying? And then maybe it's time to hire for a sales rep. But that was definitely the moment where had we been more maybe kind of attentive to the process, we would have said, hey, this is a product market for moments.

Nataraj (12:08.526) The first time I actually sort of encountered Gong was in 2018-19. I was doing a pitch deck for an Indian company trying to do software for customer support and sort of trying to do similar things of like, you know, collecting all the transcription and trying to improve the efficiency on a call center level, primarily targeting, you know, food delivery companies.

And then I was evaluating who are the bigger players internationally who are already doing some version of this. And that's my first encounter of Gong. And since then, I kept tracking of what is Gong doing. Can you talk a little bit about what types of different customers are using Gong? Because I can easily imagine all the enterprise sales organizations.

but what are the types of customers you generally can sort of categorize them into.

Eilon Reshef (13:07.115) I would say these days, obviously, the Gong has evolved from Wudang.

we started using kind of conversation intelligence and then revenue intelligence, which we added more capabilities. I'll touch on this a little bit in second. And then now kind of we call it AIOS for every teams. And obviously as the, as the sort of the capabilities were strengthened, also the types of customers that you can serve is growing. So the type of capabilities we added is pipeline management, forecasting, sense engagement, which is prospecting, coaching, enablement, these sort of things. So as you expand those suddenly more and more companies need, they might not need a whole shop bank. They might not be doing,

forecasting using software, but this might still be prospecting over software or coaching using software. So nowadays, I'd say we serve companies anywhere from a small company of like 50 people all the way up to the world's largest organizations. Five out of the top fortune 10 companies are gone customers. Cisco is one of our kind of, you

public references, they are deploying it to 20,000 sellers, which I think is the largest revenue I deployed in the world, I don't know, but obviously large scale as well. So I would say nowadays it's less, the industry we're much more diverse now, so there's financial services companies, there's healthcare companies.

of course, technology companies, even like telecommunication companies, AT &T and such. And then nowadays we also cover, people think of Gong sometimes as selling, but we really kind of try to help everybody along the customer journey for anybody who's creating pipeline prospecting in tech, it's called SDRs, selling team pre-sales, solution architects, whatnot, implementation, post sales, people responsible for retention and expansion. And then sometimes it's

Eilon Reshef (14:56.003) of the more strategic level even product managers and sort of like non revenue related. So if you sort of look at these type of personas almost every company in the world has them. We're still a lot of our business in North America just as this is where we started. Still a lot of our business maybe 50 % is still tech or tech related just with this where we started but very diverse nowadays.

Nataraj (15:17.966) What do you think about this idea that anyone can build for wipe code or stuff like that? We're sort of seeing the commoditization of writing code in that era. Getting an MVP version of Gong might be easy.

are easier than what it would be like eight years or nine years back when you started. Then we are seeing transcription software companies everywhere. There's so many of them. But my general question is how do you look at the commoditization of software in that scenario? What is the edge? What is the mode? How do you approach just building companies? You are an established player, but someone who's starting now, how would you think?

What are your general thoughts on this?

Eilon Reshef (16:08.097) Yeah, maybe I'll answer that a little bit at a zoom out level and maybe even a little bit provocatively, right? I think we are, as a universe, we're a little bit in a sort of a, you know, post-truth world. And what post-truth means also is you get a lot of incentive of just like coming up with very bold, not necessarily true claims because they get you publicity and sometimes recognition. And the press loves this because it gets them whatever call them clicks, right?

I think some of the discussion around like, we're going to vibe code everything kind of comes from.

There's the idea of, what if I just told you AI is going to make your engineers 30 % more effective? This is boring, right? Who cares about 30 % more effective, right? So now people are much more excited about talking about maybe engineers is going to go away. Maybe the PM can code. Maybe there's no need for SAS. I think these are widely exaggerated. think AI is phenomenal for engineering productivity. I don't know if it can save. I don't think anybody is saving even 50 % these days, but I think 50 % is within reach. 10 % people are getting today.

I think maybe even more. And I think it's really good for software companies. It's good for the universe because you can get more value by getting more software. I think, yes, you can definitely do LavaBull or use any other tool for prototype very, very quickly. But once you want to get like real software with all of the infrastructure, security, functionality, iterations, enterprise quality software.

Yes, it could be cheaper, but I don't see majority of functions going away. And then you also need salespeople to sell it and you need marketing people to market it. 100 % sure you can do it more effectively, but I don't think that fundamentalist change. I also don't think that organization should be bothering with.

Eilon Reshef (17:52.384) I don't know, coding their own software, they're get into the same cycle of like, gotta maintain it, I gotta change it, it's not working, who's gonna support it? Same challenge that people have had for maybe 40 years, don't know, 30 years for sure, which to me, like, doesn't make any sense,

Nataraj (18:07.468) Yeah, mean, think the people who try web coding is you can get into production, but…

once you get into production and people start asking, want this, I want that, and I want to improve this, and then you don't know what you have written, then maintenance and improvement really becomes a challenge. But that's actually probably what he said is right in terms of the post-truth one. It's easy to make an exaggerated claim and discuss it and promote it. I think it has a morality inherent built in it.

and which everyone is craving for.

Eilon Reshef (18:49.215) Yeah, I'll even say more. Now I'm going to insult a little bit my VC friends here, but you know, since, I mean, we are on record. I mean, I'm not going to beat me for it, but I sometimes tell our marketing people when they read some of those very, very bold claims. If you see them coming from VCs, for example.

mean, Gong does content marketing, right? We tell people about how sales should behave because we assume that if people read about how sales should work, eventually they're going to look and see who Gong is and then maybe they come to us as customers or as prospects. Right? Everybody else got the marketing. VC is content marketing, which is used for deal flow, right? It's basically come up with very, very bold claims about AI because that gets you the very eager entrepreneurs. So I think it should also like reverse engineer who says one into what are they kind of, are they actually writing objectively or is there a goal behind it?

all be more particular about how we interpret what's written out there in the media because that's the 2026 work, right? We can't change it, we can just be more aware of it.

Nataraj (19:47.535) Talk to me about this, you know, from conversational intelligence to this, now you call yourself as like a revenue operating system, AI revenue operating system. Like, what is the difference, you know, what are the features that make it different?

Eilon Reshef (20:03.881) Yeah, so when we started the whole notion was we're to start with analyzing a single conversation. And there's lots of stuff to be said about a specific conversation. Did you set up next steps?

Did you just talk yourself to death? The gong is probably invented to get you off measuring a talk ratio for the rep, right? Should be obvious, but people say there's a reason why we have two ears and one mouth. You should just listen more than talk. But there's so much value it can bring by just focusing on the conversation. So very quickly in the road, we said we don't want to be the experts in how to handle a specific conversation. That might be good for consumers and support over the phone. We want to help people really kind of realize their full potential in terms of being

revenue professionals and revenue organizations. So we started understanding what are the key workflows that people have within revenue organization and started rethinking about them in what I would now call AI before it was just like data and data science ways. So I'll you an example. Every revenue organization on earth, there's some sort of a cadence where somebody reviews their pipeline and decides what to do next. Sometimes it's the rep, sometimes it's a one-on-one meeting, sometimes it's a big forecast call, right? So we said, what does this process look like in an AI-centric world?

AI actually shows you which deals are more relevant than others. So it shows you, have you had a conversation, what was said in the conversation, when an event came out, you can ask a question about that deal or account or whatnot. And then it summarizes for you and it just helps you kind of do that job. So we gradually built more and more workflows. So revenue intelligence and then revenue AIOS is basically a pipeline management. And then we grew this to forecasting. What if AI can help you focus where you are, which is super important.

And then we took another key workflow, which is making people better. So we have an enablement product that says, hey, I'm going to actually help you coach the team. And of course, nowadays AI can actually coach for you to a certain degree, right? Score calls, understand the facets. And then now we're launching a trainer, AI trainer, which is going to talk with you and help you simulate the customer and coach you, right? So this is another angle. And then we looked at how do you prospect? And the idea is like, what if you can prospect to people, but actually leverage your history with the accounts and be like, yeah, AI is going to write the emails for you. It's going to do most of the…

Eilon Reshef (22:09.123) majority of the boring work for you, right? So as you look at all of those things together, if you're a revenue organization, be like, yeah, I want a single platform that does all of the things for me. I don't want reps to kind of go between systems and a different data layer that I need to sort of reconcile and then have several contracts and whatnot. then, Gong nowadays is the position where we're like, yeah, we're a single OS for everybody. So you're still gonna need like a CRM. You might still need some sort of data from somewhere and other things. It's not like, I don't think there's a single contract

in the world where you could just use that company's products and nothing else. But it is a sort of a central place where revenue professionals and leaders can do the majority of their high quality work nowadays. And of course more comings.

Nataraj (22:54.126) You talked about AI training or coaching. You also mentioned like in the starting of our conversation, sales is sort of like an art. Does the coaching really make or give you the best salesperson?

Do you see that pattern as an outcome or are you saying that if someone is 50 % effective, now we making them 80%, but really the 80 to 20 is still hard? What is your take on that?

Eilon Reshef (23:24.865) Yeah, so there's effectiveness and efficiency both contributed productivity efficiency I does this all the time perhaps you write an email and all of these things just takes takes time off Effectiveness on is more subjective in some ways because how can you prove that somebody's better Gong has been I think traditionally we've been able to show that you can move the curve

You're not going to make the excellent people, excellent plus plus. Yes, you have them around 10 % for sure, but like they're already excellent. You're never going to make the C players A players. This is not going to happen, but you can make the C player C plus, the C plus B and B is the A's kind of thing. So you see the whole curve moving and you see it pretty consistently. You just learn new skills, things they weren't aware of and just become better. I don't think you can expect everybody to be an A player. By the way, you should also replicate the A players. If you're going to try to shoot a three pointer for more.

like Steph Carey, you're just not gonna make it because you know, he shows from I don't know wherever, like weird places. You should probably, you know, you can do better than what you're doing right now, but the A, like the A plus players sometimes have such a unique pattern that you don't even wanna replicate it.

Nataraj (24:28.846) I think Gong is also the intersection of this, I would assume, like some sort of AI agent phenomenon that's happening. I've seen a lot of demos and products out there, which are sort of AI taking a call for the customer and sort of doing better or sometimes same quality output.

What is your general take on that kind of sort of like when AI is actually talking to customers and now AI agents basically taking over real sort of responsibilities? Like where are we in the curve of like an option of those kind of farm factors?

Eilon Reshef (25:11.841) We're getting close. I don't think those AI agents are going to replace a B2B setter. Most of the customers are B2B setters. There's an element of relationship. There's an element of knowledge. There's an element of just continuously understanding what's going on with your customer. AI, customer-facing AI can do a…

sometimes a good job in sort of like especially like one and done B2C phone calls, know, kind of anywhere from like a glorified IVR to just like, hey, let me qualify a little bit and sell you something. This is already happening. I think the equivalent in B2P might be inbound leads, especially ones that you don't have capacity to deal with. And the other thing is, this is kind of where I think the market will be heading is

Outdoor specific tasks from your day. So let me give you an example, right? I'm a salesperson and I want to walk my customer through my proposal, right? As an example, right?

you can send the quote-unquote agent to do this for you. If the agent's been trained enough to sort of understand what does your contract look like and trained enough to understand what the customer need, they can probably do a reasonable job in walking you through the contract, saves you like 30 minutes, and to be honest, probably the customer's gonna be happier, because now they can do it whenever they want, like 6 a.m. Pacific, you know, whatever the thing, 6 a.m., whatever, some time where the rep is not even available. So I think it's gonna be carving out those tasks, making sure that you can train

the agent to do a good job for this particular task versus everything and chop away pieces from the rep's work. I think the way we're looking at it, what's really nice about it is when we're starting to provide those things, including the trainer, the way, we train the system based on actual conversations. So we go, have a tool called AI Builder, which basically says, hey, let me look at historical patterns, identify what's working, build me something. It could be a document for humans. It could be a script for agents.

Eilon Reshef (27:10.945) and we like take based on that, which is really huge because if you started like program this from scratch, you're probably not going to get much. And then how are you going to conceive of all of the issues? It's going to take a month and a month of training. Whereas if you have all of this history, you can very quickly like iterate.

Nataraj (27:24.908) mean, it's also possible like if I'm purely dealing with customer support calls, know, hey, I'm trying to do a refund or like a very specific small problem. think it's.

much easier for someone like you who has all the training data available to create an agent targeting a specific problem. And every time a customer asks some type of question, you can route it towards that particular agent. I think that's a pretty possible scenario that I can see playing out. Do you have any thoughts of how this AI agent space will evolve? The AI agent space evolve?

Eilon Reshef (27:57.088) Big what, sorry?

Eilon Reshef (28:03.137) How will it evolve in the future? I think it will gradually take away, start from…

taking away very specific, like you and I mentioned, right? Just like start taking away those tasks that are very specific, you know, like you said, ask for refund or help me understand how my account works or maybe helping book-driving for all I know, and then gradually take more and more and more responsibilities. It might take away some of the…

maybe kind of more junior jobs, which, you know, kind of AI usually is where it shines. And then outsource pieces of the more advanced jobs. So even if you're a B2C seller, B2B seller, sorry, and you're selling very high end equipment, you might still use an AI seller as like a sidekick. Go for it. Hey, go do this for me, especially if it's certainly, you know, tasks and chatbots and whatnot for sure. But even like go talk to the customer about this particular thing and then send them off. It's not a high,

It's not a critical conversation and it's focused and it's like repeatable. So you were able to train some of the AI on it. You'd probably be able to do it with AI, right?

Nataraj (29:18.68) Do you see, so one of the patterns that I've seen is like companies that are really benefiting from.

you know, like post-HGPT sort of like the breakout, know, LLM usages, companies which have been in the play targeting a specific sector, they've already, you know, have great software. Now they can use AI to build and like sort of compound that which you sort of like fit in. Like, is that what happened when you first like saw LLMs coming out popular? I'm assuming like you were much more closer to the ecosystem. But what was your thought when like, I think in 2022 or

when our chat GPT came out, like, did it realize like, okay, now we have a really advantageous position, like what were your thoughts when that happened?

Eilon Reshef (30:05.227) So first of all, we have been using GPT 3.0 even before ChatGPD, had our own models to generate next steps. You can believe it, like we had the next steps model before and all of the things that are nowadays just like a simple prompt to say even like a, I don't know, a very, very weak model, right? So there are basically two options, right? Option one, which is the one you mentioned, which is the one we believed in is…

AI needs to be embedded in workflows, which typically tend to be kind of implemented using software. And this is kind of where real efficiency comes in. And then as an incumbent, you can come in and be like, hey, I own this coaching workflows. And in 2016, there was a manual scorecard, 2020 something, you can have like an automated scorecard or, you know, same idea with forecasting, right? had to a number and IOS is going to help you input a number or field in the CRM or whatnot, right? That was our belief and hope.

And it kind of, our thought process was this is where the world is heading. In all fairness, it was so new that we also had some percentage where we said, hey, weird stuff would happen. So yeah, I mean, you asked about vibe coding, maybe vibe coding is going to happen like in a way that I didn't think that was going to happen, but maybe, or maybe it's just going to be easier for new players to come in just because they can rethink about software in the sense of like, maybe you don't need a UI or whatever the thing is.

I think more and more it became clear that it's just like, yeah, you still need a UI, still need data, you still need a database, you still need connectivity integrations and reporting and whatnot. And AI could be an important piece or a critical piece or just a piece depending on the domain or the vertical. But you still need to have the whole shebang. If you do need to have the whole shebang, go on this position itself. think, like you mentioned for other companies in a place where we had so much IP already in this like revenue operating system that it's very, very hard for a new player to go be like, Hey, I'm going to do pipeline.

management, I'm going to do forecasting and you can't just prompt your way into this because you know when you review your book of business you don't want to see like a chatbot you want to see like visually what's happening.

Eilon Reshef (32:09.629) And the other thing that happened for Gong, which is nice, of course, is there's much more demand right now for AI. So every CEO in the world is like, hey, how can I do AI? And then ask the CRO, how can I do AI? And then they look around and they see us. But it was definitely not obvious when ChatGPT went out. were like, yeah, we're going to bet that this is the right direction. But we had no certainty that this is where the world would go.

Nataraj (32:32.194) Can you talk a little bit about fundraising and how you guys approached?

fundraising and you've been an entrepreneur before, Tegamit has been twice entrepreneur. What was your approach to fundraising? You guys have raised significant amounts of money over the course of eight years. What was your general approach and what was the fundraising journey like? Was it obvious that people like, it is generally easier for repeat entrepreneurs at least to raise the first round. What was the overall like?

thesis or approach towards fundraising.

Eilon Reshef (33:06.953) Yeah, when people ask me about advice, which generally I'm not a big fan of just asking people for advice, I usually tell them it's my biggest advice is start with the second gig. It's always much, much, much easier. You learn from all the mistakes in the first gig. But I think we did a pretty traditional route. We basically said, Hey, we need to start with a product, you know, the goal of a seed round to get you to an A round. So we need to hire this size of a team. And then get to a point where I mentioned before, like we have a decent number of customers and conviction about product market fit. So we kind of did a spreadsheet and said, Hey, we can be

$6 million, we went and raised $6 million, we bought however many people and got to a point where I think we had like $2 million of revenue and then we said, now we need an A-Round and so on and so on. There's probably some exception where I guess in the, what was it, maybe 2021-ish, when sort of valuations were at the all-time high, we basically said we don't need money.

given that the market is sort of valuing companies, we are a very high growth company at the time we still are, but like at the time it was like glowing. We basically, since the market is valuing companies of our profit very, very, in a very good way, we'll raise a little bit more money. So we have money for rainier days, which we didn't have like really rainy days, but you know, a couple of years later, obviously the market fell and it was good that we had as much money as we ever needed for.

to make mistakes, right? It's good to have spare for making mistakes. Now we're cashflow positive, so obviously we're doing nothing with the money, but I think it's always good if you can afford it and if you kind of run into an era where money is relatively cheap, buy yourself some insurance.

Nataraj (34:46.126) You've been the Chief Product Officer. What's your approach to general product leadership? Like how do you look at your own job and responsibilities as?

Eilon Reshef (34:57.889) Yeah, product is a very multifaceted role, or function in the organization. So I don't like the term CEO of a product, but it's basically saying, hey, you've got to be a little bit.

Obviously, you got to own the roadmap and what's being developed. You got to be close to engineering, you to be close to marketing, got to be close to customers and the business and whatnot, all of that stuff. At Gong, probably I, we tend to be maybe more focused on customers than anything else. So some product people tend to be close to engineering. Let's build and manage the development lifecycle. Engineers are our…

and tend to be pretty like autonomous. I mean, of course they work with product, but where I asked my team to be kind of at their A game and also how we hire is a product manager who can talk to customers and kind of reverse engineer the customer's problem. Because one of our company's motto is raving fans. We try to have customers be truly like supportive of the company, not just like customers. And for this to happen, you want to have like product managers who very close to customers. We have tons of design partners at every point in time, every feature.

we launch everything we do is like close to design partners. And if you are working close to a design partners, it's also like an insurance policy, right? You're not going to go wrong. Maybe you go slow. Maybe you're going to do stuff that's like meh, but you're not going to be like developing stuff that nobody cares about. For us, every feature we launch usually has a couple of dozen design partners at least.

So it kind of gives you confidence what you're developing actually adds value. I would say both myself and people I hire, what we hire in the team is customer centricity. Sometimes people call it customer empathy. don't necessarily like that particular term. And sometimes I err on the side of favoring this over things that are also important for product people like come up with a great solution, be a great project manager. Cause I know that if you know the customer needs and you kind of really, really internalize that, even if you can't do it

Eilon Reshef (36:52.403) job and execution and whatnot you're still going to get to where you need to be whereas if you don't know that you're just going to execute flawlessly in the wrong direction.

Nataraj (37:00.758) Yeah. Do you have any thoughts on like trends or whether it's AI or in general enterprise that you are looking at and think that will affect either Gong or in general the industry?

Eilon Reshef (37:17.417) and many, course, AI is like, you know, everybody talks about AI. think AI morphs every single job. and each person should be thinking about, you know, how should, what are the things I'm doing right now that I can do better with AI? And I think just like, again, we talked before about the hype in the media, just ignore the hype in the mini like, Hey, AI is going to replace you. is going to be like, yeah, there's not going to be no more, whatever product managers, engineers, whatnot. But do spend your time thinking where it can do your job better. So obviously for products like, I can create a prototype.

I don't need anybody's help to do a product. That's an important thing. It can help me write documents. Just please don't write generic documents and expect them to be the MRDs, PRDs of the word, right? It can help me be a body for thought process. It's really hard as a product manager to be like, yeah, I need to sort of like learn a field and iterate on a solution, right? It can help me come up with naming and terminology for things. Just don't rely on it, but it's a good thought partner. So I think every piece of your work should be, of course, influenced by.

I think the software products themselves, it's a big question. I don't think the world has yet figured out where does AI…

play out within software applications. Every vendor has got its own take. know, there's Microsoft Copilot and there's Google This and there's, of course, vendors. But I think every software is going to have, of course, some component of like any assistant within it. But even beyond this, I think this is just V1, it's like where does AI really help? And beyond obviously being the superficial chatbot layer that everybody's going to have. And I don't think the world has cracked this just yet. I think this is probably one of the more fascinating problems in

software right now, which we're gonna replace your people here.

Nataraj (38:55.854) Are you guys doing anything internally to sort of encourage the audience to leverage AI effectively?

Eilon Reshef (39:07.509) Doing lots of things. don't think anything that people haven't heard before. So on the engineering end, we actually have an evangelist team who owns this developer experience and evangelizes using tools like, of course, the cloud cards of the words.

I just heard that our bill for engineering coding has went up, I don't know, over 50 % last month, which I guess should be good, but nobody can measure the impact. So it's like for now we're just paying money and hoping it gives you effectiveness. On the product end, we're doing more of a kind of this, I guess, regular meetings where people present findings and learnings and then what have they done with AI? But I think also,

I think we kind of try to do what we preach is I would much rather bring in software that uses AI and just have it solve the job for me. So we just brought this like, it's a small startup called Bagel and they kind of do like a request management using AI. So basically if something comes up on a gone call, they bring it in. If a ticket comes in, they bring it in many, many other things. And it basically kind of use LLMs to kind of cluster it and assign it and give you some insight around what our customer is asking for now.

We don't just develop what customers are asking for. Maybe they're asking for crazy stuff, but it's really going to take away a lot of manual labor that we need to put in place to organize the data. Of course, you can look at much more data. So this is using AI, but we aren't using quote unquote AI in the sense of like, you're chatting between doing things. We just bought like an AI powered piece of software that's a pain for us. And the more people come up with these things, I would much rather have them than training people to copy and paste.

things into chat GPT or I'm hearing people are like asking people to build MCP servers and all of this stuff like an engineering job on like, just come on, just give me the software that's already been built. So we're definitely looking for things that are creating help centers more efficiently guided towards many, many kinds of pieces of software delivery, GI, those kind of things in the product.

Nataraj (41:10.67) I these are the most overhyped things, either in AI or in general right now.

Eilon Reshef (41:20.161) I think the most overhyped things is agents replacing people. Obviously, it's like agents will replace people. I've actually had a conversation with a customer today that replaced these very junior, inbound salespeople with AI. But people saying, hey, lawyers are going to get replaced. Engineers are going to get replaced. I feel like it's, again, to our point before, it's like a media thing. It's like, why would you even care? So it's like.

Let's say you have 10 engineers, why do you want to replace five of them? Just make them twice more effective. So yes, maybe it would, if you got nothing else to develop, yes, maybe you could reduce your workforce, but it's not going to be like an AI.

developer, it's going to be an AI sidekick, who's going to make every one of your developers faster. So just the thought process of replacing 100 % of your job, to me this doesn't make any sense. Just think about what is the 50 % you can replace and 50 % is quite a bit. mean, usually it takes a decade to reduce 50 % out of every job. And then if you insist on the last five…

5 % you're going to spend a lot of time on, know, in the case of engineering, it's like, how do I ensure quality in the case of product? How do I provide insights in the case of salespeople? How do I create a relationship? Whereas if you focused on just augmenting people, you're going to get much more back for the buck and leave people to do what they actually good at versus necessarily trying to take that and kind of automate, which again, to me, doesn't make any sense.

Nataraj (42:41.358) In some sense, it's similar to the self-driving problem. You can easily get 5 % efficiency and then spend a decade in solving the rest of the 10%, 5 % and 2%.

Eilon Reshef (42:51.937) Exactly, exactly, exactly. And I think it's kind of exactly mirrors that eventually yet. Obviously they are already kind of self driving kind of pilots in various cities in the United States, but it took them a couple of decades. And I think maybe for car, there's a maybe a great reason because like you either own a car, you don't own a car. There's like a big, there's a big binary thing around like, you know, is the car. Yeah. And then you can kill people. Yeah. There's so.

Nataraj (43:12.064) There's a real danger of, there's a real physical danger.

Eilon Reshef (43:17.921) The idea is exactly the same, the dynamics might be different, look, if I'm an organization, got whatever, a hundred people and they can reduce 50 % of the workload, I'm not looking to replace one individual. I'm actually trying to make everybody more effective. And yes, if I don't, if my business doesn't justify, in fact, I got nothing else to do with those people. Sure, I'm going to replace some of the people, but many, many businesses on earth get much more to do than they actually.

get to do. And probably the other thing that people maybe overlook, this may be interesting for the listeners here, I think people overlook the fact that the bar also changes with technology, right? So when I started engineering, I was programming as a kid, it was the 1970s or 80s. And the expectation was like a terminal that's going to ask you questions.

And engineering became easier and now it becomes this gooey thing. And engineering became further helpful. Now you're expecting to see animations and screens and helps and like, all this wizardry that software does today. And in essence, like writing software is not even cheaper now. You just write much better software, just more, but you could write software in four weeks in the past and you can still do it now, except it's different animals, right? I think for many, many, many domains, just the bar is going to get higher. So the domain we're in, like revenue or sales.

I think the expectations are going to be much higher from salespeople. So now we're going to expect you to actually know what the product you're selling and to come more prepared for meetings to create more detailed presentations and documents. And suddenly, yes, we've taken away 50 % of your work, but then we replace it with something that adds value to the customer versus something things they do today like filling in CRM fields, which is complete drudgery, right? It doesn't add value to anybody.

Nataraj (44:55.288) Yeah. I mean, it also means that companies like yours or your generation of like SaaS startups will know how potentially fast to attract the profitability and higher margins. Like, because if you're operating on lesser employee number, you're not really replacing humans, but you don't need more humans.

to get to the same amount of revenue and same amount of profits. I think that's one of the sort of like 100 or 12 stories. Like what will happen to that companies which already have some sort of product market fit and all you add top AI and you're still growing, you still have product market fit, you still have new customers coming in. And would that mean that you'll go IPO at a much better margin versus what you could have gone?

Eilon Reshef (45:43.06) I agree. think it's our generation can really kind of leverage those efficiencies.

And I think the expectation for companies like Gong is to reroute some of those efficiencies saving into higher growth. If you can't, you're still going to get the efficiencies. If not, or not if not, actually what all of us would want is probably higher growth. The nice thing about the market today is there's lots of demand to AI software. So for us, if we can get more efficiencies, we're going to always try to sort of route it towards fitting this demand. Of course, as long as there is like very, very high demand for AI kind of software. So, but definitely the

the factors change. So if the expectation was like, you're an engineer, you can do X, now we can do X plus 30, X plus 40, makes a big difference.

Nataraj (46:28.398) What are your favorite AI use cases that you personally sort of like use them or like interesting or unique?

Eilon Reshef (46:40.277) Yeah, personally, you know, as an executive, know, our lives are pretty boring because we don't do much. Therefore, AI can't, know, can't like, we don't have like specific tasks. I use AI a lot and I hear other executives do this as well because a lot of our work as executives is in communication. Not necessarily like form of communication, right? This like email to the company, but like, hey, coming up with the right terminology, coming up with right name, coming up with, you know, kind of organizing certain narratives around whether it's stuff we're building or…

products we're looking at and whatnot. So my main use case for AI is kind of this thought partner around positioning, wording, communication, and such. I think it's very typical for executives. I think other people may not be spending as much time in this kind of engagement.

Yeah, that's going to my main use case for sure.

Nataraj (47:33.166) What are the most exciting things that you are excited for going in the next couple of years?

Eilon Reshef (47:41.986) It's a little bit along the lines of what we discussed. So think being an AIOS is just being able to do more and shave more pieces of revenue professionals types of, for example, we're launching.

an orchestration product and what the orchestration product does is two things, automates tasks, of course. But the other thing, also like orchestrates people. So it tells them, Tuesday morning, what about this upset opportunity that you kind of forgot about? Or this customer churned three months ago, what if you reached out back to them and here's an email you can be sending them just hit the approve button, right? So the idea is like cover more and more of the revenue professionals work, both from a productivity perspective, which is what I just mentioned, as well as from an intelligence perspective, intelligence, meaning how

CRO, I want understand what's going on. I go to AI and then be like, hey, tell me what's working. Why are we not converting from stage two to stage three, which is like industry revenue jargon for steps of the sales cycle? What are we doing wrong? And in the past, these things used to take months. And the fact that you can do it in minutes, sometimes seconds, and then you can kind of as intelligence, like now I go orchestrate a change.

And the other thing is as part of this cycle is like as a leader, can do intelligence. I think the orchestration piece is also maybe underestimated because in most organizations, not just revenue organization, change management is like super, super hard. You want to do something, it takes forever, right? You got to make a decision. You got to tell everybody, blah, blah, blah, blah, takes forever. If you're using a system such as Scom, any system that kind of drives your workforce essentially.

what we're starting to do, and I think this is kind of super exciting as the next phase, we're gonna let companies kind of vibe code their business in the sense of like, you know, tell me what are the reps doing in Australia that's different than New Zealand? And be like, okay, so that's what they should be doing. You click a button, suddenly it rolls out different plays to the people in New Zealand.

Eilon Reshef (49:42.306) And then otherwise it would have taken you months to understand what's going on, months to activate. And you're also limited in your capacity because you have to sort of train the people. There's a whole motion around how many initiatives can you run in a month. So the idea of like AI learning from what works and what doesn't and the idea that you as the operator can vibe code your.

I call it vibe code. It's not vibe code. It's like use AI to sort of build those plays, right? But you don't have to sort of like go to a UI and just like be like, do this, do this, do this. And then with the click of a button, roll them out to the field. And it's not just for revenue for me, it's revenue is exciting. Makes this whole optimization cycle super, super fast. And then it can help to also like hyper personalize it because now we can roll a specific play to Chicago versus Denver, right? Why would you do this? I don't know. But you just cannot do this today just because of scale, right?

There's no way for you to go to do that.

Nataraj (50:34.552) think what you're basically saying is some version of every platform, be it GOM or some other product. think every product is now sort of almost…

logically forced to have an orchestration plus some version of app platform created on their platform because now it's easier and it allows a lot more people to create things that are useful and share with other people. think that's what you're basically, you know, sort of like a meta trend that is happening. And you'd also already see across like Microsoft products and some other AI products that everyone has some version of orchestration that sort of like obviously needed because that sort of empowers the product.

ecosystem a lot.

Eilon Reshef (51:18.301) Exactly. And if you have not every company does, but we have a revenue graph, which is like really, really deep. If you have data around historical performance and what works and what doesn't, not just numbers, numbers usually kind of hard, but like if the data layer is rich enough and you can have AI mine it for, for what is the right play, then you actually can close the loop. Right. So now it's like AI gives you ideas and like what you could be doing. And then AI helps you do the things using an orchestration layer and then it executes it using the orchestration layer. It really transforms the way you kind of.

changes happen. And a lot of why people quote about hire us as software vendors is to drive change, right? I mean, nobody buys software to the same things in this same way that you can just use pen and paper, right? And the ability to kind of iterate very, very quickly on those kind of changes is, think, you, people don't even realize it's possible, but certainly it's going to take probably a few years until people just like execute this. And I think it's as transformative as the idea of AI agents.

but again, maybe not as sexy to the press, but probably as transformative from a business perspective.

Nataraj (52:21.12) In some sense, actually, white coding is needed within the enterprise platforms rather than like this individual, like make a static website, which is actually very impressive, but also not useful because we need more stuff on the enterprise layer where you can wipe code stuff. mean, which actually becomes much more powerful because, we spend more of our time at work and not like, making hobby blogs.

in some sense.

Eilon Reshef (52:51.489) 100%. And then you're integrating with the data, you're integrating into a workflow. It all comes because otherwise people are still going to be copying pasting stuff from people. You're not going to get efficiencies if you keep copying pasting stuff. It's just not going to happen.

Nataraj (53:03.714) I think that's one of the reasons why actually the adoption of AI tooling is actually lower on enterprise side because most of the tooling is being done on the B2C hobby prosumer side and that has not still like created in the enterprise ecosystem. think we still need that shift happen when actually people get more productive.

Eilon Reshef (53:24.929) 100%. I think prosumer is where it shines right now, which is phenomenal. I'm sure you use it. We, use it. Everybody uses it. But I think AI has been a little bit slow to sort of.

people to get their hands around how do I connect it to my data sources, how do I connect my data flows to my existing applications and so on. But obviously it's coming. Look, I'm super excited about this whole idea of AI centric or first applications. Because to me, the infrastructure layers, I don't want to say software, but API, know, so like OpenAI and Cloud and all of these companies are like a drop in, I mean, Google, all of these guys, they're going to, I mean, obviously the LLM kind of infrastructure is.

I don't want to say sold, but it's being handled by so many companies and it's a lot of white space on the layer on top of it. And I think for Gong it's a big opportunity, but for many, many other companies there, I think there's still a huge opportunity.

Nataraj (54:14.744) think that's a good note to end our conversation. Thanks, Ilan. Thanks for coming on the show and sharing all the interesting things that you guys are doing at GAN.

Eilon Reshef (54:23.852) Thank you for hosting me. Have a good conversation.