Wispr Flow: The AI Siri & Alexa Couldn’t Build | AI Siri | Alexa | Young Turks Reloaded | CNBC TV18

You started building when you were in 5th grade.

Couple of 12th graders there who were building something, and I went up and asked them, "Hey, what are you guys doing?" And they said, "Oh, you're too young, you won't get it." And no one says that to me.

So let's talk about the first big venture.

And we built the world's first device that went from thought to text. We thought it was a massive moonshot, but It was built at the wrong time.

After having got investors on board for the hardware business, to then say, okay, it's a pause here, or it's actually a full stop.

All they said was congrats. Congrats for realizing before you basically spent all of our money. And when we launched Whisper, it was unlike anybody was expecting.

If I were to ask you what the plan is—

I want this to be the most loved AI product in the world. ChatGPT. Siri, Alexa. This is June 2024. And turns out they all sucked.

So what made you think that you, a startup in the Valley with no experience in consumer tech, would be able to crack the behavior change?

We were building something that at that time actually sounded like a ridiculous idea. If in 2024, Shereen, I came and told you like, hey, we're building this dictation app. You'd be like, why? I just feel life is short.

You're not even 30.

Is still short.

How do you then create a moat around yourself?

Right before I was coming to this meeting, if I was messaging somebody like, hey, I'm going to meet Shereen, it knows how your name is spelled because it knows I'm going to meet you.

What was the environment like? I mean, to have two overachievers grow up Simultaneously.

I never imagined myself ever having a job. And so ever since I was 10—

Never? Not once?

Never. As an 11-year-old, I'd write the name of my company on a piece of paper for like whatever I did, and I would end it with Inc. My dad was like, Tanay, you cannot add Inc. to the name of the company. You actually need to incorporate in the US. And I was like, no, no, let's incorporate it. And he was like, in India you'll get private limited. And I was like, I don't like private limited. I want Inc.

So are you as thrilled close to your Jarvis moment.

If 10-year-old Tanay could look at what I'm doing right now, he'd be very proud.

For over a century, the keyboard has been our default interface with technology, but in the AI era, that's beginning to change as voice emerges as a faster, more intuitive interface. Hello and welcome to Young Turks Reloaded. I'm Shereen Bhan. From a childhood dream of building a real-life Jarvis to a bold pivot to software, Tanay Kothari is now betting big on voice as the future of computing. Founded in 2021, his company Whisper has built Flow as a voice-to-text AI that doesn't just transcribe, but it structures, edits, and refines your thoughts, all in real time and across any app. After a dramatic reset where he cut his team from 40 to 5, Whisper doubled down on Flow, and the results are showing. The company is now growing over 40% month on month, with the average user typing 72% of their characters with Flow across nearly 70 apps, cutting their daily typing time in nearly half. Backed by $81 million in funding and strong product-market fit, Whisper Flow is betting on a future where voice isn't just a feature, but the primary interface for computing. So are we on the cusp of a post-keyboard world? And what does it take to build the operating system for voice in the AI era? Joining me now is Tanay Kothari, the CEO of Whisperflow. Tanay, thanks very much for joining us here on Young Turks Reloaded. Let me start by asking you about how you got started. And it's not a 5 or 10-year journey. You started when you were all of what, 8 or 9?

I started way back, yeah. And also, thanks for having me.

It's great to have you.

Yeah. So I started building things back in around 2008, primarily because I was in 5th grade. And there were some people—

That just throws me off. You started building when you were in 5th grade.

I was. And so this was when I was in my school's computer lab. I went to DPS in Delhi. And there were a couple of 12th graders there who were building something. And I went up and asked them, hey, what are you guys doing? And they said, "Oh, you're too young, you won't get it." And no one says that to me. So I went back home and that was the first time I pulled an all-nighter to teach myself how to code. I didn't have a laptop, so I stole my mom's laptop from the living room. And the entire night I spent learning, building, and then right before my mom was supposed to wake me up, I made sure her laptop was charged back up to the exact same level, put it in the same place, and then pretended to sleep. And that basically defined what the next now 18 years of my life looked like.

And how long could you keep this a secret from your parents?

I don't think they ever figured out until I told them right before I was going to college.

So you were doing, you were doing alternate all-nighters without your parents ever finding out?

They had no idea.

Okay. And so from that point where you taught yourself to code to the start of your first entrepreneurial venture, which also happened while you were in school.

Yeah, so throughout middle school and high school, I built a number of different apps which gave me more than anything really good intuition on how to build products for consumers that they'll love and that changes habit. And so when I was 14 or 15, this is in 10th grade, I got in touch with Arjun from Investopad. He now runs Good Capital, which is a great VC firm in India.

Yeah, they have been on Young Turks. Yeah.

Fantastic. Yeah. So Investopad at that time was one of the first startup incubators. And so he met me through a hackathon. He invited me to come and be his intern for the summer. And then one week later, lo and behold, we both decided to start a company together.

Okay.

And so that was basically the time where I grew from, you know, an introverted software engineer into somebody who was a lot more articulate, a lot more open, because the thing that he taught me was, Tanay, you can't build a company by just being behind a screen all the time, because so much of it is about human relationships.

Yeah.

And he's one of the people I just think of as an older brother who really helped me kind of get out from being this shy little kid. And that was the first company that we worked on for a couple of years right before I went to college in the US.

Yes. Let's talk a little bit about your journey in Stanford because that kind of became the ground for you to do what you're now doing. And you had the opportunity to work with people like Andrew Ng, the founder of Coursera. What was that like?

It was magical. So Andrew started Coursera, which is also how I got to both get deeper into machine learning and learn more about Stanford because he has this course online that I started watching in 9th grade. So for me, he was one of my idols, right? So to to go there and have a chance to work with him was a dream come true. And the thing I learned about him when I was there was I understood why he was one of the most famous AI researchers out there is because he was relentless and more ambitious than anyone I had seen. And so when I was working with him in the AI lab, me and other people, we would just spend 60 hours a week just doing AI research in the lab. And remember, we have classes, assignments, extracurriculars, all of that as well.

This is all happening on the side.

This is all happening on the side.

Yeah.

And then this became my main focus. And it was like, imagine a small room. There's 25 of us sitting shoulder to shoulder, all of us who were kind of highly ambitious undergrads for most part, who he said, I'm going to teach you how to do research and publish papers in under 3 months. And we're going to publish in some of the top journals. Now, the reality was, We did not publish everything in 3 months, but every single one of us solved a really big problem in the healthcare AI space that nobody had ever touched before. And so the one that I had worked on was this model that, you know, in hospitals when patients are just lying down, what often happens is they have a blood clot that forms and it gets stuck in their pulmonary arteries, which are the arteries in your lungs. And that's called a pulmonary embolism. And that causes 10% of hospital deaths. And the thing is you can figure it out with a CT scan, but the radiologists are always backed up. So you get your results in 3 days, 4 days, but fatality happens on day 1.

Yeah.

And so the thing we built would actually look through the CT scans, figure out if there's blood clots, and actually surface that to radiologists that, "Hey, this is something you should look at first 'cause this is high risk." And then it was deployed at the Stanford Hospital and a couple of other hospitals as well and actually started saving lives. And so that, again, because he pushed us so hard, we were able to build it, deploy it to such a good degree. And one of the things I'm still really proud of.

You know, so I would imagine that after a success like that and that early on in your life, you weren't tempted to go down that route of addressing more health tech or doing an entrepreneurial venture in health tech?

I was, but that was not my calling just yet because the thing I realized was what really gave me joy was the time I'd spent building products here in India. And the biggest part of it was I'm walking down the street and I see somebody use a product that I've built. And there's a big smile on the person's face, that matters to me more than anything else. And so when I was thinking about what I wanted to do after college and where I eventually wanted to land, that is what drove me more than, more than anything else. And so probably health tech, I'll come back to this because there's some really important problems I do want to work on, but they often require a lot more resources. And if you really want to make a big dent, you just need to be more influential overall. So not something that is a good idea for your first big venture.

You know, so let's talk about the first big venture, which is not quite the venture that you started or thought of when you actually got started. From hardware pivoting to software because you realized that you know, the hardware part of the journey wasn't working as per your expectations. And it would perhaps be the business that you would need to kill before you moved to the real business that you're in today. What was that like? What got you started and how difficult was it to kill that to be where you are today?

So I'll tell you the story of Whisper, right? So it's 2021. And I have just sold my first company that I started right out of college. And I'm working at this patent company and I get a call from Sahaj, who is now my co-founder.

Right.

But at that time he was my best friend in college, my roommate all through undergrad. And he said, Tanay, I want to start a company. Now, he was the actual research kind, and I never thought he would want to start a company. So I was like, you know what? He's the smartest guy I know and I cannot let this opportunity go. So we started chatting about what were some of the biggest things in the world that we saw that were important problems to solve, but nobody else was working on them. And so at this time, GPT-3 had come out. Yeah. Not too many people knew about it. This is way before ChatGPT. And when we started diving into it, the thing that was clear to me is you can now finally talk to your computers just like you talk to other people. And in the next few years, we are all going to be doing that. 'Cause voice is the modality that's most natural for humans.

Yeah.

And so in a world that is changing, in a world where you're gonna have AI and large language models be a part of your everyday life, how do you interact with your devices? How does that need to change? And so we thought, okay, you know what? OpenAI is of course gonna build the best voice model and any one of these large labs. And then the Siris and Alexas of the world are gonna build an incredible voice assistant.

And they're already here.

And they're already here. And so the thing we need to build is how do you use voice around other people? How do you have privacy? How do you not disturb them, right? Because we all work in open offices. We all are around other humans all the time. And so we decided to solve the problem of silent speech.

Okay.

And the eventual goal has always been the same, which is I wanted to build Jarvis. I wanted to build a voice assistant that's with you 24/7, just gets you, is able to take care of all the grunt work in your life so you as a person can just be focused and present. And one day when you go down on the streets, you don't see everybody hunched over their phones like this. They're looking up. They might have something in their ear. Ah. But because technology fades into the background, they don't need to be staring into their phones any longer. And so that was the core thesis that was true then. It's still true now. But back to the problem of silent speech, right? So the thing that we built, 3 years, 40 people, mostly PhDs in neuroscience, signal processing, electrical, mechanical, and we built the world's first device that went from thought to text.

Okay.

It was a Bluetooth earpiece-like thing that you could put on and you could speak silently in your head and it would convert that.

What do you mean speak silently in your head?

Imagine saying hello in your head.

Okay.

That is what it would be able to pick.

So literally not saying anything out loud.

There was no microphone on this thing, nothing out loud. And we actually got this to work. And it was one of the most magical pieces of technology. I have some videos. We can even like show that as a part of this. But then the thing that happened was we connected it with ChatGPT, Siri, Alexa. This is June 2024. And turns out they all sucked as a part of feeling like a natural voice interface. And so we realized we needed to build something that goes from our mental rambles into something that was structured and ready to send. And that made us build the operating system for this hardware, and we called it Flow. And so at this point, while we were building that, The thing that we realized was, remember how I said the first thing that you need to build is voice input that just works? It's reliable.

Yeah.

Then and only then can you build an assistant on top of that so people trust it to do things. And then finally you can build hardware so people can use it everywhere. 3 years later, June 2024, nobody had built part 1. And so that was the big realization that we had is if I give this product to people, there is nothing that they will use it with. And so this hardware product is going to flop. At this time, we had a lot of capital in the bank. Our investors were really happy. The technology was working. We thought it was a massive moonshot. It was like 5% chance we're actually able to build this, but it was built at the wrong time. And so Sahaj and I took the hard decision of killing the hardware business and focusing fully on Flow, which was this little dictation product that we had built. It had just about 100 users. And the thing that I knew was these 100 people loved it more than anything else, and it changed the way they work. You would come into our office and you'd see everybody speaking into these little mics and using it for everything. And basically nobody in our office used the keyboard anymore. So we did the pivot in August of '24. And basically at this point we went from a deep tech brain-computer interface company to now building consumer software. And so to get the team back up to speed and make the transition happen as quickly as possible, I told the team, gun to your head, we need to launch this product in 8 weeks.

Huh.

In 8 weeks we built a fully functioning production level software. We built a full marketing motion from scratch. And that was the big launch that we first did in August— October of 2024. And the rest is history. But I think overall—

very short history though.

Short history. Yeah. That was one of the hardest times of my life. Those, those 2 months when we decided to kill it because it was something that I'd spent days and nights working on for 3 years.

3 years. Yeah.

And I was really close to every single person on my team, right? Like personally hired every single one of them. I was close with a lot of their families and they had done incredible work. So it was not even that they weren't doing good work and that's why we were letting them go. They were doing incredible work. And the only reason was we didn't believe this was the right thing to build, which made the decision a lot harder because I couldn't put it on anybody else. And I think for me, when I think about building businesses, your products change, your markets change, but the people stay. And so that has always been my number one focus and how much I personally invest in the people. It was a hard lesson that taught me a lot of mistakes that I had made and just made very, very sure that I don't make them again.

You know, I think acknowledging the fact that you could have done things differently is the start to actually breaking a pattern that you perhaps may have found yourself in and, and then course-correcting. But I want to go back to, to what you said about you were perhaps ahead of the curve. And we've seen this play out for many businesses during the course of the, of the last 20, 30-plus years. Why do you feel you're in the right place at the right time today with Flow?

So the key thing I realized, and this is for, for consumer businesses, is the best consumer businesses change user behavior. And the biggest thing about changing user behavior is people don't like to change. Generally, you like to keep doing things the way you've been doing them your whole life. And so if you really want to push somebody to make that change, you can only change one thing at a time, right? For example, If you are familiar with this company Humane, they built an AI pin that you could put on, you could talk to it, but what you had to do was you had to get a new SIM. A lot of things that you were doing on your phone you had to do in this new device. It was all voice first, and this was a lot of change that you needed people to make. And so with Whisper, our pitch was really simple. It's like, hey, you have a life built around WhatsApp, Outlook, ChatGPT, whatever else. Don't change any of that. Just keep using the same things. All I want you to do is at least try to speak instead of using your keyboard. Just try the simplest thing. Try it a couple of times. And just that little thing started to build that first layer of habit change. The next layer of habit change that now makes sense on top of that is, okay, if you're now comfortable speaking to your computers and dictating everything, you are now comfortable also giving it commands to do things on your behalf. And so that becomes the second layer of behavior change that we build. And then finally, what comes after that is I can say, Shereen, this thing you love using on your phone and laptop, imagine if you didn't even have to take your phone out of your pocket. And that is the final thing where now you have the habit of voice to both write, do things, and now I'm just telling you to use a different device for it. So each of these is a single step in habit change. And whenever we're building anything new, I would go and validate that with a core group of people to see if this hypothesis is true and this habit change is really a simple step.

You know, as you pointed out, habits are hard to break. Now, what, what is the data telling you in terms of uses, in terms of adoption? Yeah, I mean, eventually your, your plan would be to make sure that there are no keyboards any longer. I mean, that would be the ideal situation. In your head as part of your plan. But I mean, what are we actually talking about in reality in terms of adoption today?

What we actually see is when a person starts using Whisper, 90% of our users have never used voice before, right? So this is their first time entering into voice, mostly because voice has not been something that is high quality enough where you could use it in a day-to-day life. So they start off and the first thing that they find is one place where they like using Whisper. It could be to respond to all of their emails, all their WhatsApp messages. It could be for all of their, say, AI prompting or Wipe coding or anything else like that they're doing. And about at the 4-week mark, they hit this realization of, why don't I use this everywhere? And you see this in the data where at the end of month 1, 20% of everything they do on their computer is with Flow, 80% is with their keyboard. And 5 months in, it flips where about 20% is with their keyboard and 75 to 80% is with WhisperFlow. And so we do tell people like, hey, this is something you need to invest in a bit, but this is the way the future is gonna look like. And people find it worthwhile enough to, to make that change. And so this is both a boon and curse of building a product like this because it asks a good amount from people. But what we deliver to them is an enormous amount of delight, saving time in their life and getting ready for what the next 20 years is going to look like.

So what is the next 20 years going to look like? You know, I mean, there's all kinds of scenarios being painted out. On what AI is going to do to the world, what AI is going to do to jobs. But what does AI do to the world that you want to inhabit and how do you envision it?

The key thing that I always focus on is, and this is how we, how we choose, how I choose to build what products we focus on, is say I would sit with you and I just want to learn what all do you do in your life that feels like grunt work, that feels like work that you would rather not do. So you have time to do what you genuinely love. Maybe that's talking to people. Maybe that's diving deep into a particular field and learning more about it. And my goal is I want to give you as much time as possible to go and do that for your creative, deeper pursuits, for your family and friends. And so you take a look at all the parts that feel like grunt work, and our goal is to go and let AI take care of all of them for you. And so in this way, what we're using it for is increasing human agency and augmenting people, not replacing them, because that, yes, it's good for the P&Ls of some businesses, but that would be a net negative for the world if we treat all of AI like that.

So let's talk a little bit about how you managed to get investors on board. You've got $81 million in the bank at this point in time to do what you like with it, to try and accelerate the growth of Whisperflow. Now, what was that process like after having got investors on board for the hardware business to then say, okay, it's a pause here or it's actually a full stop here and we're moving to a completely different animal, a completely different beast. What was that process like?

It was a test of faith in a lot of ways. So for the initial part of the business, we call it Whisper 1.0, which is the hardware business. We got some of the best investors in Silicon Valley, firms like NEA, 8VC, a lot of incredible founders like the founders of Dropbox and Whoop and Warby Parker and a lot of these companies. And when we decided to pivot the company on a Monday, I remember I'd booked in-person meetings with every one of them. And Saj and I went and told them our reasoning for why we were thinking of shutting down the hardware business and switching over to software. I was expecting them to ask to potentially do a recap, maybe like sell their shares, maybe find a way to like shut down the company fully and have us start again separately. But all they said was congrats. Congrats for realizing before you basically spent all of our money that the thing you were spending it on was not the right thing to spend that capital on. And we will support you for whatever you want to do because at the end of the day, we invested in the two of you as founders. And you just assume that things are going to change. And this was interestingly what we heard from every single one of our investors. And so, and this is even though we were building something that at that time actually sounded like a ridiculous idea. If in 2024, Shereen, I came and told you like, hey, we're building this dictation app, you'd be like, why? It's on every device for free. It's just the button right there. Citi has it, Microsoft has it, Google has it. Like, it makes no sense for you to tackle this market. It's 20 years old. That was what everybody thought. And so while they were supporting us, I don't think most people really believed in this thing that we were building. And my only reaction was like, give us a couple of months, right? Because there was something that I saw, which was, yes, it exists everywhere, but nobody uses it. And that blows my mind because this should be the most intuitive way to interact with technology. And the other thing they didn't believe in was we were a deep tech hardware company, right? So in their minds, we're like these researchers who are great with building this. For a consumer software company, it's a completely different mindset. You need to be great at marketing, you need to be great at building products, shipping fast, or trading fast. Completely different DNA of a company. And so that was the second thing that they were unsure about, about how we would do. And the great thing was they were very upfront about both of these things with us, and I was like, cool, this makes sense. And again, give me a couple of months and I will show you. And so when that happened and when we launched Whisper, uh, it was unlike anybody was expecting. I think that first launch day hit about 10 million impressions. We had all the who's who of Silicon Valley who had started to use the product. And then it just was a rocket ship after that. And so for the next couple of rounds that we did, it was just preempts from people who invested in the company because they believed in not just where we were at that point, but how we were thinking about building the business and where this could potentially grow if Sahaj and I keep doing what we're doing for the next few years.

So let me go back to what you just said. What blew your mind was the fact that you had access to using voice to do pretty much what you wanted to do, whether it was Siri, it was Alexa. I mean, voice notes on all of our phones and so on and so forth. And you're right, we don't use it as much as you would like us to or as much as they would have liked us to, each of these companies that had put these features on. So what made you think that you, a startup in the Valley with no experience in consumer tech, would be able to crack the behavior change?

It was a really simple insight. And the really simple insight told me that everybody so far has approached the problem in the wrong way. And so the thing is, if you want to, say, get subtitles for a YouTube video, right? The voice models that are out there today, great. If a couple of words are wrong, it's fine. If, you know, it's not properly punctuated, it's also fine. And you want every single word word for word written there. But when you as a person are using voice, that is exactly not what you want. Because guess what? The way you speak is very different than the way you write. Like usually when you're speaking, you just say one long 2-minute sentence. And when you're writing, it's structured paragraphs and all. People are extremely particular about how their writing looks, where commas go, where full stops go, what is punctuated, like capitalized versus not. And this differs per person and for every single different application and person that they talk to. And so our key insight was people don't use it because they have an extremely high bar for the output quality, and no one building these technologies realizes that. And it's not because these people aren't smart— like, these are some of the smartest researchers in the world who are building this— it's because of how these companies are structured. So if you look at any one of these companies, you have the product team, right? They talk to the people, then they tell the engineering team what to build. Yeah, the engineering team builds it, and then there is an R&D team that is encapsulated away from everything else.

Yeah.

And they're like, "Hey, we need a voice model that does voice-to-text." And so then they would go and build it, but you don't actually know what the person on the other end wants. Mm-hmm. And so that makes it that the models that are finally built are very different from human behavior.

Mm-hmm.

And so the key thing we changed within Whisper is we had our researchers both build everything inside the product Right, so they're shipping and building every single week, and they're also the ones talking to the users. And that made us extremely grounded in human behavior and what people want. And that level of detail is what made Whisper extremely delightful, 'cause now when anybody uses it, they're like, "Oh my God, this is perfect.

I don't need to change a thing on what Whisper has written down." How did you get to that point, you know, to take 2-minute rambling sentences and put them into structure and, you know, with punctuation and so on and so forth. Take us through what happened at the backend to be able to get to that outcome.

It's actually crazy. Enormous amount of attention to detail. I'll give you one example.

Yeah.

So commas. It's one thing people are extremely particular about and they— people use commas in different ways. And when we looked at the data, Turns out there are 15 different reasons why a person would like a comma, and that differs per person. Some people like commas in certain places, some don't. And so, for example, what Whisper does, as you use it and as you correct its mistakes, it learns the ways you like to use your commas. And so the moment you try any other tool, it just feels wrong, and you would have no idea why, but it's just like, This is just not how I write.

Yeah.

And this is one thing with commas, and we do this for every single thing. So for example, on my WhatsApp, it's learned that I don't end my messages with full stops because people think I'm mad. My dad likes grammatically correct messages, so he puts full stops at the end of everything. And Whisper just works differently for me and him, but it works exactly like both of us would like it to work. Other things also, for example, is a lot of times I realized that when you hear somebody speak, unless you know the context, sometimes it's like kind of confusing, like, hey, what did you say?

Yeah.

And so you kind of need to know the situation you're in to be able to understand that properly. And so that was the second thing that we focused on. How can we learn about the person, people who matter to them, to be able to figure out how to write their name? So for example, right before I was coming to this meeting, if I was messaging somebody like, hey, I'm going to meet Shereen, It knows how your name is spelled because it knows I'm going to meet you. And so it's going to spell your name the right way, even though there's a number of different ways to spell a name that is pronounced as Shirin. And so these were all little, little details that we spent a lot of time focusing on. And the final thing was, this was a really funny one. So when you use any otherwise dictation tool, it shows the words as it's dictating.

Yeah.

Right?

Yeah.

Now there's one problem with it, which is the words keep changing a little bit as you say the next thing, which is fine, but we found that people feel that it is less trustworthy because they're like, "Oh, you're just constantly changing your mind. How do I trust you actually know anything of what I'm talking about?" And so with Whisper, the thing we do is once you're done speaking, it shows the whole thing at the end. So one, it builds trust, and two, as a person, you're not reading while you're speaking.

Yeah.

And that actually feels so effortless to use. And so a lot of these specific insights, which just were very different than how any other company had done it before, were how we started optimizing it. And all of these things that I'm telling you are obvious in hindsight, but those were the key things that we were realizing, like, okay, if I as a person want to really trust this, like, this is what we would need to do. And then actually solving all of this was a really hard engineering problem. But I was very happy to take that up because that would mean that if we are able to actually solve it, we build something magical.

You know, talking about building something magical, and I want to understand from you, in the world that we live in where things are changing so quickly and big tech has pretty much sort of, you know, taken up all the large spaces available as far as consumer tech is concerned. Does that worry you that perhaps today they're not— they haven't been so far interested in getting this absolutely right? But maybe now if they decide to focus on this, how do you then create a moat around yourself?

So this was a top of mind question throughout, right? Because in the space we're in, the biggest people who I think of as existential risks are the OpenAIs and Apples of the world. We're building for the layer between you as a person and everything you want to do with technology. And that is the space that generally any large tech company wants to come into as well. I would say this has increased for both Whisper and every single software company out there from all the developments that we're starting to see happen in the last 2 months. Which I think has caused a general existential panic among all the software companies that are out there. And so this gets me to a couple of key realizations about software as a whole. So the one big thing that has changed is software today is practically a commodity business. Commodity business, just like shoes or shirts, because anybody can build software in a way that you could not until 2 months ago. What that means is you need to start building these companies like commodity companies. And so what makes Nike really special? It is the brand that they've built. It is the emotions that they evoke in people. It is the space that they own in people's minds. And so building the brand is one of the biggest things that you need to do today as a software company if you really want to succeed. Because anybody can copy your technology in a few, a few days.

It's not so much about the software anymore. It's more about the story.

It's a lot more about the story. So your brand is one thing that is still very unique, and that is how commodity businesses have survived and thrived around the world. And the second thing that is even more important is when you, when you look at a lot of these companies what you realize is they build for the mass market. They build something that a lot of people in the world can use, but that means that there are a lot of communities that are just left behind who aren't really well served. And those are the markets that we really want to focus on. And so my goal whenever I'm building for a specific market is I want to be the best product for people here. I want them to feel like This is the tool that is built for people like us. So now, for example, when we were entering into just doing a lot more in India, well, first thing we did is we started out Android first, right? Because Android is a huge part of India. The second thing we did, we supported all major Indian languages, and Whisper today is the number one voice model for every single one of them. So we're talking like Malayalam, Tamil, Telugu, Marathi, Punjabi, Hindi and a lot more.

And Hinglish.

And we added Hinglish. And that is huge because that is actually from a machine learning standpoint, an unsolved problem until today. There are zero companies in the world that have solved the problem of people switching between two languages.

And have you now successfully been able to do that?

Now we have, because I told my teams like, okay guys, we are launching in India in March, and before the launch we need Hinglish to work, otherwise I'm going to be embarrassed taking this product to India and my parents won't be able to use it. And so that both push and prioritization as a company, right? I told my ML team like building Hinglish is the number one priority that you have. Made them actually crack it. And now Whisper is the best and also the only model that supports Hinglish because my goal is, remember I was telling you, I don't want people to think because the biggest thing in India is we speak so many languages but we are just forced to do everything in English, which just adds an extra load. You're likely translating a lot in your head all the time. And I just wanted to meet people where they are. And so building Hinglish was great and we're getting so much love How good is the Hinglish on Whisper? It is better than most other languages, and we keep improving it time and again, but it's better than what you would get from any other model everywhere. The key metric that we track is what percentage of your messages are perfect. And so this is a key metric that we care about. And so, for example, if you dictate a 3-paragraph email and one comma is off, that to us is a score of 0. We have failed because it's not perfect. So that's the bar we hold ourselves to. And with Whisper, only 8% of the messages that Whisper outputs need any editing. So 92% of the time it's outputting something that you don't even have to read before pressing send. Hinglish is at a 12% mark right now, a little worse than English, but still better than everything else. And as we grow more in India, that number is going to drop down.

Well, you know, I'm going to have to try the Hinglish on Whisper.

I want you to.

Yeah. And see where it goes and how it works. So what's the grand plan from here? You know, now, now that you've got into this space that you want to, you've hit the sort of sweet spot. What's the plan from here?

The plan is by the end of 2026, I want to get to a point where I have something in people's hands that gets them to take their phone out of their pocket 30% less. This is the original thing that we started off with. And I think finally we are there from both a technology perspective, a company perspective, capital perspective that we can actually go and build that, which is going to require a lot of work on both the model side and also getting back to the hardware side. So you can expect to see some— So that's the plan, getting back to hardware? Whisper hardware coming soon. It's going to look very different than what we started off with.

What's the form factor likely to be?

I will let you know. Maybe the next time you come in June to our offices.

No, we look forward to that. But so how many languages can you do today? How many markets are you operational in? Which are the markets that you intend to prioritize?

So Whisper is today running in 105 languages. It is being used by people in over 160 countries. And right now, US is our largest market. India is the second largest market. And then we have basically a lot of places in Europe, Asia, North America, South America, a few countries in Africa where Whisper is used by people. And we really think about our global expansion having a phase-by-phase basis. So we're starting off right now with going really deep in India and going deep in the United Kingdom. And then after that, we're going to expand to more countries in Europe and APAC.

So when you talk about going really deep in India, what is the Whispher story going to be for India?

A lot of people are surprised when I tell them like we're focusing on India as a market because the notion in people's mind is like, oh, like Indian consumers don't pay. This is like, are you just getting more users here or what? But the thing I realized is, number one, India is actually one of the fastest adopters of AI in the world. Number two, this is fundamentally a voice-first country. We call instead of texting. It's the biggest usage of the voice features on every single application. And this is just how overall the country is.

And also more inclusive.

It's way more inclusive. And what we started to see with a lot of the early usage is if people feel a lot of value that they derive from your product, people pay. So our conversion rates in India are actually very similar to the conversion rates in the US.

Okay. Because how much are you priced in India?

So in India it's ₹300. Okay. In the US it's $12. And so of course we wanted to make sure it's in line with some of the other products and all that, that people pay for similar to the Netflixes, ChatGPTs of the world. And that has now seen a massive spike of growth across right now all of the major cities and starting to get into other segments as well. For example, like people in the age group of 50 to 80 who, for example, like did not grow up with a keyboard, who aren't extremely tech savvy. There's a lot of people who would type on their phone like this. And for them, voice is just delightful. So there are so many people actually who, once they find Whisper, like people in our generation, they would go and then get their parents to start using it. So we're seeing expansion across all of these different parts in the country.

So what is the demographic cohort that you intend to focus on in India? I mean, you give us a great example of the use cases as far as the 50-plus, you know, population is concerned, is there a specific cohort that you intend to focus on here in India?

So the one that we're starting off with right now is what you would think of as the white-collar professionals. These are your engineers, managers, designers, salespeople, and these organizations. So these, these departments within, within companies, and this is across startups, large companies like the telcos and banks, all the way to small SMBs that are coming up. But these are people who would generally, for example, have good value for time. And so my proxy for that is this is the kind of person who probably uses Uber or Zomato for their convenience because their time matters to them. Because for people like that, getting a couple of extra hours means that they can do more, whether that's more work or have time for, for friends and family. And so that's the area where we have seen immense amount of pickup.

So enterprise adoption as well?

Enterprise adoption is starting to happen in India. We have a phenomenal enterprise motion in the US. In India, I would say it's picking up. It's on the early stage of the hockey stick, and our consumer growth is usually in people who work at these companies. So If you think about all the top startups in the country, the CEOs and CTOs of all these companies use Whisper. If you think about all the top VC firms, all the partners there use Whisper in their day-to-day life. If you think about all the startups that are up and coming, like 10 to 50 people, their entire teams are using Whisper. And so this is usually like the wave of adoption that we see. And then now for the remainder of the year, a big focus is breaking into and getting into a lot of the large telcos, banks, all of these places and make it the default way in which work is done.

So how many users altogether between all of the markets that you operate in?

We have a few million users and 5,000 enterprises.

5,000 enterprises largely in the US?

5,000 enterprises, I would say 90% of them are based in the US, and this includes some of the trillion-dollar companies, a lot of the Fortune 500s, top startups like Notion and a few more that I can't name. And then in Europe as well, where we have companies like Klarna that have Whisper deployed wall-to-wall across their entire company.

So what does it mean in terms of revenue? What does it mean in terms of profitability?

So the great thing with how we've built the company overall is these— I care a lot about building the fundamentals of the business well so it's ready to scale. So Whisper runs on a close to 90% gross margin. It's between 82 to 91 based on the month. It is also one where we have incredible retention. So our consumer retention after 1 year is about 70%.

Okay.

And that for reference, for the average consumer app on mobile is about in the single digits. And best in class is about 40 to 50% for some of the top, top AI companies. And within enterprises, we see a lot of land and expand. So you'd go into a company, we would give it to, say, the engineering team. And then before you know it, 6 months in, half the company is using it. So usually we see every single one of our enterprise accounts grow by 60% every quarter, and that really helps with the revenue growth. So business is growing at about 40% month over month. It has been for the last 18 months. It gets harder to maintain that every single month with larger and larger volumes, but that also keeps the team on their toes to learn new growth and GTM strategies to keep the pace going. And that is honestly a big part of the challenge that excites me a lot.

So over the next 5 years, if I were to ask you what the plan is or how you really see the company change and evolve, what do you have on your vision board?

I want this to be the most loved AI product in the world. I want this to be something that everybody knows about.

More than ChatGPT, more than Claude from Anthropic, you know, the most loved AI Product?

The most loved AI product in the world.

That's a big claim.

That is a— it's a goal.

It's a goal.

It's a goal.

Yeah.

And everything I do, that is the thing that gets me up in the morning because maybe there's a 1%— maybe there's a less than 1% chance that we get there, but it's not zero. And as a person, you know, I just feel life is short. Life is short to do—

You're not even 30.

It's still short. And if you're doing something, just go all out and see how far you can take it. Because if you don't get there, that's fine. That would have happened anyway if I didn't shoot for the moon.

Yeah, absolutely. And many more moonshots to you and to your team. The big lessons, the 5 top lessons as a startup founder that you'd like to share with other startup founders, uh, who are on this journey and dealing with their own dilemmas?

I'll say the first one is you don't start a big company in your first go. Every time you're starting something, what you really need to focus on is a very specific problem and then just be the best person in the world at solving that. 'Cause if you are able to do that and build that muscle, building a big company is just doing that small step repeatedly again and again for years and years. And that's how every single big company started. A lot of times what people do is they try to take this big audacious challenge and either it's wrong or they're not resourced enough to be able to go and build that. Second thing I would say is don't listen to what your users say, judge them on their actions, because that is going to teach you a lot more about user behavior than their words do. So for example, for the first 500 Whisper users, I onboarded them on a Zoom call, and I wasn't looking at what they were doing on the screen. I was looking at their eyes and their expressions on where are they delighted, where are they confused, because those are the things I'm gonna go and change. Third thing I would say is The people you bring on in the company matter a lot because what the right people do is they reduce stress from your life. They're able to take a big chunk of what would otherwise have been on your plate and own that for you. And at the end of the day, the thing you need to focus on is how do I get 5x, 10x, 20x, 100x more leverage in doing the thing that I'm doing? And that both means getting the best people in And also spending a lot of time with them to teach them how you think. So they have that same mental model that they go in and approach things on a day-by-day basis. So you can multiply yourself a number of times. That was 3?

2 more to go.

2 more to go. You're putting me on the spot.

We can stop at 3 as well. The top 3 is good.

Top 3 is good. Let me see. And the final thing I would say is a lot of people when they think about building a startup, you forget that you're actually building a business. And so the key things like your user retention, how sustainable and high quality your revenue is, what is the gross profit you're making with every single customer that you're bringing on. Are actually really important metrics because sometimes you might say this is not important, I just want to grow right now. But it makes growing so much easier when you have the base set up well because there's fewer things you're stressed about and a lot more that you can just pour fire on.

So are you as close to your Jarvis moment that you dreamed of when you were 9 and 10 and decided to get into the world of technology?

You know, if 10-year-old Tanay Kothari could look at what I'm doing right now, he'd be very proud.

He'd be very proud.

He'd be very proud.

But your parents should be super proud. I mean, it's not just you. You've got a younger brother who's also raised funding and has a startup, and he's what, all of 21?

He is 21. Yeah, we have a 7-year age gap.

And so how— what was the environment like? I mean, to have two overachievers grow up simultaneously in one household. I mean, I should be interviewing your parents. My parents actually.

You sure? They're fantastic. So both my parents were initially working in companies. My mom was a VP at a bank. My dad was a CFO at a company and then MD at Accenture. And then during my teens, both of them decided to start their own companies. So I saw them going on their entrepreneurial journey by themselves, build things from scratch. I saw them have 18-hour workdays And I saw them be extremely delighted and high energy even after all of that, just because they were doing something that they loved. And for me, I never imagined myself ever having a job. And so ever since I was 10—

Never, not once.

Never. I was always, I want to start a company, I want to build this. I used to, as an 11-year-old, write the name of my company on a piece of paper for like whatever I did. And I would end it with Inc. My dad was like, Tanay, you cannot add Inc. to the name of the company. You actually need to incorporate in the US. And I was like, no, no, let's incorporate it. And he was like, in India you'll get Private Limited. And I was like, I don't like Private Limited. I want Inc. And so this was just like me as a 10-year-old and what I always, always wanted to do, which is like, you know, funny in hindsight if you just imagine this like little kid running around. But there was just no other way. And so my brother also, whenever I was building something, Ever since he was 6 years old, I would just wake him up in the middle of the night and be like, "Yo, you want to see something cool?" And I would show him what I build. I taught him how to code when he was that old because I just wanted a coding buddy.

Yeah.

And so he also grew up basically building apps ever since he was 6 or 7 years old. And he got a lot more into design. So he's like the more creative one between the two of us.

Well, I wish him the very best of luck as well. So have you given up the keyboard altogether?

I have—

honest answer, honest answer, honest answer, hand on your heart.

I use it for some things, but I'll tell you the biggest thing that has happened and, and changed in my life is when you as a person are trying to put down an idea, right? Maybe you're, you're sending a message to somebody, maybe you are writing a blog or a script. There is a part of your mind that's the creative part that's generating the idea and going out But there's this other part that's saying like, oh, how do I spell this word? Comma or a colon, or oh, backspace, backspace, backspace. And when you just delete this part away and you're just focused on the idea, your mind just feels so free. And that is what I feel with Whisper. That's what everybody else feels with Whisper. Hey, for most things in life, you don't need your keyboard. You need it if you want to do like final touch-ups, editing, all of that. So like, cool. Do that, that's fine. But 95% things in life don't need that.

Well, I hope you continue to enjoy your freedom. And thanks very much, Tanay, for stopping in and talking to us here on Young Turks about the Whisper journey. And we wish you the very best of luck with your plans for India as well as all the global expansion that you're looking at.

Thank you, Shereen. This is fantastic.

Thanks very much for being here. With that, it is time for us to wrap up this edition of Young Turks Reloaded. As always, send us your feedback and your comments. We always like to hear from you. We'll see you again next week with another young entrepreneur. Till then, goodbye and many thanks for watching.