Subscribe
Share
Share
Embed
Conversations between the venture investors and operators at Dell Technologies Capital and the people who are building what's next in enterprise technologies.
DTC (00:00)
I think the founders and the executive team at Zapier, early this year, so around January, early February. So this was two months after Jad GPT came out, decided that AI was here to stay. AI is powerful. It's only going to get more powerful and we better embrace AI. So we had what we called Code Red, which was a directive for every one of the company to start using AI to make their workday lives easier, faster, better. That was one director. And the other director was to quickly figure out how we can.
I use AI to deliver a better experience to users. Welcome to the DTC podcast, a series of conversations between the investors and operators at Dell Technologies Capital and the people who are building what's next in enterprise technologies. In this episode, recorded at the end of 2023, DTC investor Radhika Malik talks with Ankit Garandas about his career transformation from being a biomedical engineer by training to an exited tech startup founder.
And then now his role as a staff engineer leading AI related projects at Zapier. Hey, Ankit. Great to have you here. Thanks for spending time with the DTC team today. Maybe just kick us off by sharing who you are, a little bit of your background, and we can take it from there. Hey, thanks, Radhika. Thanks for having me on this podcast. Really excited to share some of my journey with you and the rest of the DTC listeners.
My background. So my name is Ankit. I'm a staff engineer at Zapier, which is a workflow automation company. The company is completely remote. And one unique thing about Zapier is that it's been completely remote since 2012, since it was founded. And I used to live in the Bay Area. I now live in Denver. We took advantage of work and my wife's job opportunity and relocated here and love it. Most recently, before Zapier, I was a startup founder. I founded a company called Intersect Labs.
And eventually, Innercic Labs got folded into Zapier. For people's context, Ankit and I have known each other almost 15 years now, spent several nights in college dealing with problem sets. And we were similar majors, electrical engineering, computer science. He was a couple of years ahead of me, so gave me some words of wisdom right from 15 years ago and where we're at today. From those days in college over at MIT,
At that point in time, I think the idea was to go into research related roles and that's what you did post school, right? So what took you from that initial thinking to becoming a founder and now doing the things that you're doing today? Yeah, for sure. Yes, I was really gunning for to go into the academia, but more as a researcher. And so right after school, I joined a lab at MIT. I did some algorithm development for medical devices there. And then,
After that, I had a stint at Mass General Hospital, one of the largest hospitals in the New England area, where I worked as a researcher at the sleep lab and wrote algorithms to process EEG data, which is data recorded from the brain surface while patients are sleeping. And soon after that, I felt this urge to move into the industry because I wanted something more faster -paced. The output of academia was mostly papers, and I wanted to have something that I could build at user studios.
And so then I joined a medical device startup called CBM and that job brought me out to California. I worked there for almost two years or two and a half years and really enjoyed it. I got to dabble in everything from designing the electrodes that would deliver a small current to this set of organs called carotid body in the neck to actually, you know, working on clinical trials and.
overseeing them in Poland. So I had a blast working on that. And then I had the itch to switch again, I wanted to work in something that was not as regulated because even though I was building something that users could use, the timelines were long. There was a lot of paperwork, there was a lot of approvals needed, which is a good thing obviously, but I decided I was not at the stage in my career where I wanted to have that big of a feedback loop. I wanted tighter feedback loops. And so,
I sought to work at a more traditional tech company, which is still somewhat related to the human body. So I joined a company called Athos. Athos was a wearable technology company. We made garments that measured muscle activity in real time. So it was mostly used by athletes, by physical therapists, by people who worked out at the gym regularly. And again, had a blast there, worked on everything from the actual design of the garment electrodes to writing algorithms that ran on the firmware to.
writing algorithms that ran in the cloud. And then, until then all my job changes were mostly aspirational. I wanted to switch from academia to industry. I wanted to switch from regulated to non -regulated. The next switch was more practical or logistical in nature. My wife was starting a residency program at UCSD in San Diego. And so we were going to San Diego from the Bay Area and I decided it was time for a fresh start. So I ended up consulting for about a year.
I consulted mostly for small companies that needed help with their algorithm development or data processing. And then in working with these companies, I found the need for some platform that allowed non -tactical users to easily analyze their data and more specifically make predictions or make forecasts using their data. And so that's when I started this company called Intersect Labs. Our first product was, it was called Machine Learning in Three Clicks. It...
enabled users to upload their historical data, select which column they wanted to be able to predict, and then hit create model. And based on that, our software would create a machine learning model for them, deploy it on the cloud. And then from there on, users could make predictions both using the web UI or using the API. We took that company to Y Combinator. We raised about $1 .6 billion in the seed round, pivoted a couple of times, but...
Oh yeah, that was my journey from college, academia to starting a company. Fascinating story. I guess one of the audiences we're hoping this podcast caters to is the founder, Curious. We call it, you know, people who are either in academia or in startups or scale -ups who are thinking about starting their own company at some point as well. And obviously you had a meandering path, but I've left you pretty interesting directions.
For those people, what part of the founding experience really called out to you and what has really resonated with you the most? Yeah, I think it's the impact, right? And that was a common story throughout. If you tease apart the story, you'll realize that when I wanted to switch from academia to industry, it was because I wanted to have a particular kind of impact. When I wanted to switch from regulated devices to unregulated devices, it was, again, the kind of impact that I was hoping for. And similarly, founding a startup was all about that impact, right?
you can make a ton of impact, obviously, working at a smaller company or a bigger company. But, you know, that impact is 10X or 100X or even 1000X when you start your own company, because, you know, the company lives or dies by you. You know, the code you write, the conversations you have with customers, how you run the company internally, the conversations you have with your employees, with your investors, they all shape the future of the company. And so I really wanted to have that kind of an impact. And because I had...
an interesting idea that seemed to cater to a need in the market, I decided to act on it. Yeah. It seems like it was both the impact and then you were just looking for a faster paced environment at every point in time as well. And how was the Y Combinator experience for you as a first time founder? It was amazing. I really enjoyed it. I would definitely do it again if I were a first time founder. Until then, I was really a novice at running companies at...
raising money or even talking to customers. Until then, I'd been an engineer. I had not really been exposed to talking to users, figuring out what the business side of things are going to be, and Y Combinator really helped shape some of that aspect for me. I'm really grateful for that 10 -week journey. I had a really good time. That makes a ton of sense. And then you obviously went through a few pivots as well through that journey at Intersect Labs. So...
What was the journey through some of those pivots? And obviously you're talking to a ton of users, you're building fast. One of my combinators real advantages is really their focus on making you shift and iterate pretty quickly. We see that in a lot of YC founders that emphasis on doing that. So did that help in that process? And then how did you think about the pivots and how was that journey for you as you tried to find product market fit and that early product development and validation?
You know, the whole YC ethos of shipping fast, building fast was incredibly useful. And that's something I carry to this day, even now that I'm not at my own startup. Yeah, I think in terms of Pivots, we started the company around middle to end of 2018, went through YC in summer of 2019, and then started sales earnestly around that time, basically summer of 2019. And we were doing pretty well, we were growing and then six months later, the pandemic hit.
February of 2020 was the pandemic time. And suddenly, all discretionary budgets were just canceled at every single of our customers. And we were squarely in the enterprise sales game. Our starting prices at the time were about $25 ,000. And then most contracts were a little bit higher. And so it required some amount of approvals. So users couldn't just come in, log in, put down the credit cards, and start using it, unfortunately.
And because of the nature of those kinds of sales cycles, we either saw a lot of those potential deals just get lost or people started telling us, okay, you know what, let's, you know, let's talk again in a few months. A good example of that is the U S tennis association. We had a signed contract with them. They'd use it for a month or they'd just onboarded in early February of 2020. And then when the pandemic hit people stopped playing tennis. Eventually people realized that you could go out and exercise, but.
Early on in the pandemic, it was all about just sitting at home. And so we were getting hit pretty hard. And so we decided, hey, can we build something that lets users just come into the website and start using it and pay small amounts of money for. And the product we found the need for was a more data analytics product. So think about an operations person who wants to upload their spreadsheet and have some sort of a join of that spreadsheet with some data in Salesforce.
And then they want to do some analysis, maybe some pivot tables, maybe some aggregations, maybe some filtering, and then output the final report into a Google Sheet. To do that is cumbersome. And to do that repeatedly is extremely hard. And so we basically built a product that allowed non -technical users, mostly business users, to build and automate these data processes. And we started working on it in October of 2020 and came out with it soon after.
I think we were out with it by the end of the year of 2020. Yeah, that was our main pivot. And rationale for that was a lot of business driven decisions. But the actual pivot and what we built came from users. We talked to a lot of users. And we just decided, OK, this is what users wanted. Yeah, that's also interesting. We're also in a similar macro environment today with enterprise budgets, being a lot slower, especially for non -core projects. And I'm pretty sure.
entrepreneurs are going through a similar journey right now, which is, hey, should we continue to believe in our product and this is what we want to build and this is what we want to execute or is it time to pivot and think about a new product? It's also interesting that by pivoting the product, you also probably had to think about a different go -to -market motion. I guess that was the ethos of doing that as well, right? Trying to move away from top -down enterprise sales to more bottom -up product net related stuff.
So how was that retooling your go -to -market motion as well, in addition to pivoting the product to do that? Yeah, it was fine. We were a small team at the time, there was three or four of us, and I was the only go -to -market salesperson at the company. And listeners cannot see this, but I just put air quotes up there next to me saying I was the salesperson or the GTM person. And so it was not that hard, right? I think at earlier stages when there's few of you,
a lot of the GTM just comes from how the product is. And so we were able to quickly adapt to it. We built a product that was geared towards bottom -self adoption, and we were able to execute on that pretty well. Yeah. I'm guessing that was a nice segue into the next chapter of your life because Zapier, the company that ended up acquiring Intersect Labs, was all about product -led growth. That's right. A lot of sub stuff. So how was that journey also, deciding, all right,
I'm going to now do this at part of a bigger platform versus trying to go the startup route. Yeah. And in a lot of that was governed by both, you know, or all three, you know, some of the macro conditions that were prevailing at the time. And then, you know, how the company was doing at the time and then things that were happening in my personal life, you know, in terms of the macro, the economy was definitely recovering. The companies were raising money, but budgets were still not completely loose. We're talking about middle to end of 2021.
So, you know, we were still facing some headwinds in terms of adoption and sales. The company was not yet on a path of becoming a unicorn in two or three years by any means. We were growing, we were growing pretty well, but we were not three -axing year over year, which is what you should be doing if you want to become a unicorn. On the personal side, we were expecting a baby. And so, you know, we had a huddle with the rest of our team and said, okay, you know.
let's see if we can get acquired by a bigger company and continue our journey there and have a shot at this again. And unfortunately the team responded pretty shavably. And we actually, around that time, we got some inbound from another company. And so that's when we started having these conversations internally. And eventually, we ended up talking to Zapier, which is also a YC company. And really enjoyed talking to them. There was one other company that was in the mix until the very end, but ultimately,
Decided to go with Zapier because Zapier seemed more, they had this division called Zapier Labs at the time. And it was definitely more, you know, focused on innovation, on building new products. And I really enjoyed talking to the founders, specifically Mike Knope and Brian Helmick. Mike was the president at the time, Brian is the CTO. And so eventually we ended up joining Zapier and it's a decision that, you know, I am really grateful for. It's been, it's been an amazing journey at Zapier.
Yeah, no, that's great. And it's also a decision that founders go through after they decide, all right, we're going to try to go do this in a bigger platform and finding that fit, as you said, is also a hard process. It's critical. Yeah. And I'm glad you did that diligence and you've been here for a bit. So it seems like the right decision, but tell us more about your role at ZapierNabs and now where you're at today and what you're doing a little bit more. Yeah. So...
I joined Zapier Labs in March of 2022. So it's been about a year and eight months, a year and nine months. Our charter was very simple. Go find a need for something that users have, that specifically our users have, Zapier users, and then build it. And the only requirement is that it should have the capability of becoming a big business in five years.
And so we tinkered quite a bit, talked to a lot of users. And that's something, by the way, I really respect about the Zapier founders and how that's trickled down to the rest of the company. Talking to users is highly encouraged and not just highly encouraged, it's required. So if you show up to meetings and say, hey, I think we should build this or I think this is how we should design this feature. If you don't have evidence backing, this is what the users want. And not through surveys, not through analyzing data on Looker.
purely through conversations you had with customers. If you don't have that kind of data, then you don't have any round to stand downs. And that's something I really, really admire about the company. And so yeah, I started at Zapper Labs. We talked to a bunch of users. We found the new core, a couple of different things, so started prototyping it, presented it to the company. They were all AI related. So this was, you know, March of 2020 was about six or seven months before, about nine months before, Chad GPT came out.
So AI was not as ubiquitous in the world. It was getting ubiquitous in the tech world. I think we had GPT -3 at the time, and the model was great. You mostly had to use it using APIs. It was a playground, but it was not as good as it is today. Definitely not as functional. And so we started building some products using the AI capabilities of OpenAI. Got a lot of support from OpenAI, who at the time were really encouraging companies to adopt LLMs.
Around October, which was about a month before Chad GPT came out, Zapier decided to invest a little bit more in AI stuff. And so two of the founders and a bunch of us came together to start an IT at Zapier. And we ended up again, building a lot of different products. Some of those products got folded into other products. So for example, there was a bot builder that enabled you to turn your prompt into a user facing app. That was an RIG related product that
allowed you to query or ask questions of your data. And since then, I now work in this division called New Products, which is the natural evolution of what Labs was. And I specifically work on the newest product in New Products. It's not been launched yet. We're trying to figure out if we can build something that's more developer -facing. So traditionally, Zapier has been used by non -developers, mostly by business users.
And now we're at a point where we're trying to figure out if we can build something that's more developer -facing. Yeah. That's some very interesting projects. I guess on the AI side, you said Zapier's ethos is all about talking to users and using those insights to inform the products. And again, as you said, March 2022, it wasn't AI. It was a buzzword, but it wasn't as busy as it is today, Happy, after a year of chat GPT.
So were users asking you for AI related stuff or was it they were asking you for stuff and you're like, hey, this technology actually might be ready enough now that we could use it to solve some of the things that users are asking us for? Yeah, the latter. I mean, most of our users had barely heard of AI. They heard of AI in like, you know, in as much as they maybe read a few tweets here and there or read a few headlines here and there. Most people didn't know the powers of AI or LLMs at the time. LLMs was definitely not a word that most people knew.
And so users are not asking us for AI things, right? Users were giving us their problems. And we decided that AI was a good technology to solve those problems. And if I can have a minor segue here, this is where that famous quote by Henry Ford comes in, where he says, if I asked users what I wanted, I'd be breeding fast horses, not building cars. And a lot of entrepreneurs, a lot of bigger companies,
use that code as evidence to suggest that you should not be talking to your users because you know exactly what they want. But in reality, if Henry Ford asked his users and they said they wanted faster horses, their problem was that it was taking them a long time to get from point A to point B. And sure, the solution they were asking for was faster horses, but the problem was getting from point A to point B faster. And Henry Ford figured out that a way to solve that problem was to build cars. And I think that holds true for almost everything.
users will tell you what their problems are and you got to figure out how to build those problems. But the problems have to come from users. Yeah, absolutely. And it's also interesting, as you said, the company's also been to embrace this new technology. And NFRIs are still trying to figure out how to do that at this point. But it looks like Zapier's, the folks, everybody at Zapier has gone headfirst into trying to see.
Can we use it internally for our stuff? Can we use it in some of our products for our users also? So tell us about some of the projects going on for both internal and external folks within Zapier and what the different pillars there are of the AI stuff going on. Yeah, for sure. I think the founders and the executive team at Zapier early this year, so around January or late January, early February.
So this was two months after Jad GPT came out, decided that AI was here to stay. AI is powerful, it's only going to get more powerful and we better embrace AI. So we had what we called Code Red, which was a directive for every one of the company to start using AI to make their workday lives easier, faster, better. And this goes back to one of our...
one of the core tenants of Zapier, which is called build the robot, not be the robot. So we're always encouraged to automate stuff. And so yeah, that was one director. And the other director was to quickly figure out how we can use AI to deliver a better experience to users. And out of that came a few different products or features. One of them is this way to go from idea to Zap in minutes. You basically go on to Zapier and traditionally you build a Zapier, a Zap, which is workflow, things like, you know,
get triggered every time there's a new email. And then that's a building block. And then you can add the next building block, which is populate a row in a spreadsheet. So with AI, we have a feature that lets you describe that workflow in words. And Zapier will build a Zap for you that you can then add it. Similarly, we have a bunch of our other products, like the AI Chatbot, which allows users to build user -facing bots. A lot of our small business users are
building these bots to enable them to embed some sort of Q &A bot on their website that users can come in and ask questions of, and it just gives the answer. We have other products like, newer products like tables, interfaces. Tables is our superpower spreadsheet product, and then interfaces is a website builder that users can use to build user -facing sites. And both of them have a lot of superpowers picked in.
As Zapier has embraced AI really well, everyone has encouraged to get chat GPT accounts. I think last I checked the usage and I checked a few months ago, so it's higher since, but the usage of AI and specifically GPT at Zapier is over 50%. And that's, you know, across the board, not just engineering products, design teams, but also our support teams heavily use AI to make their lives easier. Our sales teams use AI. Everyone, basically everyone.
uses and is highly encouraged to use AI. Unlike some companies where the director early on was, don't use AI because if you find out use AI to automate your jobs, we will fire you because that's cheating. Zapier was quite the opposite and we were industry leaders there. And then the other thing that happened from this embracing AI was our partnership with OpenAI. We were the launch partners for a couple of OpenAI products. Yeah. So it's great to see Zapier going really headfirst into it.
And you've hinted the fact that Zapier is a very customer -centric company, very user -focused. And so now that fast forward a year and a half, a year and eight months, AI has become ubiquitous and everybody's talking about LLMs and foundation models. How has that changed today? How are people asking you for AI -related stuff today? And has it lived up to people's expectations so far?
Now that users know what's possible, especially users that play around with ChatGPT. And a lot of our users, while they're not technical, they're very tech forward. So ChatGPT is the perfect product for these kinds of people because they don't know how write code, but they know how to embrace technology. And so a lot of our users started asking for a way to build chatbots. And initially, that was a feature in interfaces. And then it got so successful.
that they decided to split the feature, turn it into a product, dedicate a full team behind it. And so that's the one example of what users were asking for that we were able to deliver. Another is being able to build workflows or Zaps from natural languages. That's also something that users were wanting and we built that. In terms of whether we're meeting customer demands or we're satisfying customers, I think Pack will never do that.
I think aspirationally, you can only get asymptotically closer and closer, but I think we're pretty close. Both things to some amazing work by the folks at Zapier and then also the underlying technology itself. So we're getting super close, but there's always edge cases where things don't work as well. But fortunately, users are very understanding. Yeah. And then we need entrepreneurs and innovators like you to think about newer use cases that... That's right.
users don't directly ask for as well. That's right. But how do you also think about success in an AI project also? There's a lot of enterprises going through POCs at this point, trying to figure out what the ROI is going to be. So were there conversations within the company to say, hey, is this project a success? What should we really move forward versus not going that route?
Yeah, I mean, for user tracing things, it's very simple. It's all about adoption and revenue, right? If users like it. And early on, it's not even about adoption or revenue. It's more art than science at very early stages. It's all about, hey, are we building something that customers tell us they like? And so going back to chatbot, that was a great example. Early on, it was a feature that customers really liked. And then eventually, in terms of usage metrics, it got up there.
which is when the team decided to split it out and turn it into its own product. I think it really depends, right? I think the ROC calculations come in, especially when you're spending a lot of money to build something small that moves the needle a little bit. I think in Zapier's case, we had some projects like that, but a lot of our stuff was user -facing. There was going to be a stop change in how users interact with Zapier. And so...
There's always RSC calculations, but especially this year, I think the team has been pretty good about saying, okay, experiment, experiment, experiment, right? Don't worry about costs right now because whatever you learn will be great, right? And I think it's been paying off really handsomely for us. Yeah, that was great. And for you, how do you think about making sure your entrepreneurial scratch keeps, you keep scratching that itch, I guess, even at a 10 year old company and -
the parallels between innovating at a startup sub 10 people to a large company, but still trying to be at the forefront of innovation. Yeah. I mean, in many ways, I still get this crash that itch on a daily basis. Like I said, I'm working on this newer product. So for about a month, month and a half, all we were doing, and when I said we, it was a team of two of us. All we were doing was talking to users, talking to customers, prototyping things and Figma.
prodiving things in VS Code. And then we got to a point where we were reasonably confident. We knew what we wanted to build. And so now we're at a point where we're building it. We're putting it in front of users very soon. I get to write code. I get to talk to users. I get to think a lot about what the business would look like in one year, in five years. And so in many ways, I do get to scratch that itch on our databases. I am extremely grateful for that. The one great thing about this is that while I
do get to scratch that itch. I also don't have to stay up all night trying to figure out if we'll be payroll next month, or if we'll hit our short -term targets. We always have targets, obviously, for any project, but we're not on a very tight timeline right now. Are there parts of being a founder that you actively miss?
Yeah, definitely. I think that whole impacting is definitely something I miss. I do get to have a large impact at Zapier and then hopefully with the product that I'm building right now, if it takes off, then it'll be an amazingly large impact because Zapier has a lot of brand recognition. But having that agency, having that impact is something I miss. And then I think there's something to be said about that pressure. I just said it's...
it's good not to have that pressure, but I do miss that pressure as well. It does bring out the best in summer class. You're obviously enjoying yourself at Zapier, but if the stars align personally, would you consider starting another company? Yeah, I think so. I think in a few years, I would definitely consider starting a new company, especially if the right idea came by, if the right business need came by, I would definitely consider it. I think I'll be...
I'll be banking on a lot of experience that I had from starting Intersect Labs and then seeing how a well -run startup like Zapier runs. I'll be using a lot of that. Yeah. So those experiences, again, going back to the Founder Curious folks, many of whom would be first -time founders as now or a few years from now, a second -time founder, what would be some lessons that you want to share with folks?
Yeah, I think this has been something that's been said, you know, a million times by other people and probably about 10 times by me on this podcast, but just leaning into that user discovery, right? I did a lot of that at Intersect and then when I came to Zappi, I realized that I could have done so much more. And then if a 10 year old company can continue to do that, a startup founder should always, you know, bank on users, not their knowledge of users. So that's number one.
I think the other thing is be a little more mindful of how the company needs to grow, especially around funding and stuff like that. I don't regret any of the funds or any of the money I raised during the first journey, but I think from that experience, I'll be a little more mindful of raising right when I absolutely need to. I'm in a fortunate position now. We've built some savings.
If I were to quit my job and start a startup in a couple of years, we probably won't have to rely on VC money to pay my salary, at least for a few months. And I wouldn't rush into raising money right away. I would first talk to users, build something, try to make money out of it. And then when it got to a point where more money would fuel more growth, that's when I would raise VC money. Because as great as VC money is, it does put you on a timeline and it does put you on...
on this path to not allowing you to start from scratch if you want it to. Right. It does limit you in how much you can experiment. And so that's what I would do for myself. Obviously, you know, every founder is different. Every founder's life is different. Every founder's company is different. If you're building a large language model yourself and need to hire a bunch of GPUs for a few months, yeah, sure. Go ahead and raise some PC money. You need that. Right. But for a lot of problems out there, maybe you can just...
you know, experiment without raising VC money and, and, you know, actually raise revenue, which is, you the best signal in terms of the philosophy of business. So yeah, that's what I would do. So raise money to service demand that you already have, versus trying to figure out if there's actually demand or not. Would that be one way to say? Well, kind of not just demand, right? So raise money when money is a limiting factor.
Right. So in my example of building LLMs, if you need to raise GPUs, money is kind of a limiting factor. And you don't know whether there's a demand for it yet. You think you do, but there's no real way to get there until you build those models. And there's no real way to get there until you raise money. So, you know, sure, go raise money. Right. So yeah, I think if I had to put it betterly, you know, raise money when money is a limiting factor. Yeah. This is a great conversation on Git. I think we talked...
about a lot of different stuff. So really appreciate your time and also really appreciate the candid insights. And I'm sure the people who are listening as well, hopefully got some interesting nuggets out of it. So thanks so much again. Thanks for having me. I had a blast. I'll always have a blast when I talk to you, Adhika, and had a blast again. Awesome. Thanks for listening to the DTC podcast. If you like what you heard, you know what to do. Like, share, and subscribe wherever you get your podcasts.