Building a Profitable AI Company: Krish Ramineni, CEO of Fireflies.ai, on Virality, AI Agents and Scaling

0:00Hello, and welcome to another episode of Riding Unicorns. Today, we're delighted to be joined by Krish Ramineni, co-founder and CEO of Fireflies.ai. Now, Krish has done an amazing job at building a profitable AI company, which is maybe a little bit rare in the current market. Yeah, we want to see more profitable AI companies. But also, there will be a lot of people listening that will have come across the Fireflies product, as a lot of people would have had a Firefly's Notetaker added to a call that they've been on at work. So it's great to have you on, Krish. Thank you so much for doing this.

0:34Let's just start by talking a little bit about the early days. You did raise a few funding rounds, but you haven't for a while now because you've been so profitable. So tell us about some of the early days and building the product and the thesis, because we're all kind of used to AI Notetakers now, but obviously you were creating a new category. So tell us about the start of the journey. For us, fundraising was actually a very difficult process in the early days. The smallest check we could get at the time, or any check for that matter, was a$25 ,000 check from a small angel accelerator. And we use that money and had it last for over a year in San Francisco, one of the most expensive cities.

1:19It's not that we didn't want to raise money in the beginning, we just couldn't. And then after a certain point, we said, maybe we're not built for raising venture capital, because everything I would hear is about people over promising founders saying they're going to do this and that. And there's a big tech crunch PR release. And then six months later, the company goes bust. So I saw many cycles of that. I saw people that started around the same time as I did, that would go on raise tons of money, and then completely go into oblivion. So what we did believed, though, was whether we like it or not, whether this works out or not, we fully committed to the idea of working for ourselves and building a startup.

1:57And before we even had a real product, we said, this is what we're going to do. As long as we can make$10 ,000 a month, that'll pay our rent, pay our servers, pay our expenses, we can be ramen profitable. There's a great essay by Paul Graham from Y Combinator about as long as you can get to ramen profitability, where you cannot be killed, you are in control of your destiny. So I really took that to heart back in the day. And I said, whatever we build, as long as we can self-sustain ourselves, this is going to be great. Once we started building these products, we iterated on five, six different products, all with the same singular thesis, which is, there is knowledge buried inside conversations, and we should be able to use AI to unlock that knowledge.

2:37Back then, it was contrarian because AI wasn't quite good. The natural language processing wasn't good. So there was a lot of skepticism around the technology. So we were doing AI 10 years before it was a thing. And as we started building, though, we felt like if we can also consumerize the AI, democratize it so that everyday people can use it, not just PhDs in a lab, that would be the real win for us. So we picked a novel problem at the time, which was, what do people do every day at work, which is have meetings, whether you're a salesperson, a recruiter, internal knowledge worker, manager, it doesn't matter.

3:10And there is so much knowledge buried inside those conversations, we talk almost three to four X faster than we type. So it made a lot of sense to go and capture those conversations. Back then, just recording a meeting and saving it to the cloud was very expensive. When we came out with our product, we said, we're going to give you unlimited recording. You can capture all these meetings. We'll transcribe it. You can search through it. And then later we incorporated the note taking element into it. After we did our beta, after we got a couple hundred customers, 500 or to a thousand users, we said, okay, we have something.

3:44People are using this like crazy. They're using it more and more each week. We're going to run out of money at this rate. So this is a good problem to have. That's when we went to VCs. That's when we went to Canaan Partners and a few other VCs and ended up raising that seed round, which was around$4.5 million all in. After that, we ended up launching live out of beta. And then that was right around January 2020. Something else happened in January 2020, which was COVID. All of a sudden, all of work went remote. Everyone was on video calls, Zoom calls, Teams calls. And we just rode that wave like no other.

4:19That just accelerated our growth at that time. And we started seeing a lot of adoption. We started monetizing after our first year. By the end of the year in 2020, we started having conversations with investors on a potential Series A. When we raised that$5 million seed, we said we're never going to raise again. But there was so much uncertainty at the time in the market. We said, OK, let's just do one more round. We'll do the Series A. After that, we're not going to raise again. We're going to use this as like a backup, like safety. And I'm very proud to say since raising that Series A, we never touched that money.

4:54And we were able to scale and eventually get to profitability in 2023. And since then, the last three years, it's been a wild journey. So we've been growing like a venture-backed company. but also a profitable AI company. And a lot of that was, yes, one of it is timing, two, it's being in the right place at the right time, and then the technology getting significantly better. And we basically rode the wave long enough until AI caught up. Awesome. It's an amazing story. And you stay true to your principles throughout. Can you give some examples of some of the things that you actually did to try and keep your costs down?

5:30At the time, we were trying to be very first principles. Elon has this thing about if you're trying to build a rocket, you got to go and test every assumption. If someone says this is how things have always been done, I don't think that's a good enough reason. So we'd go back and figure out, okay, how do we engineer our costs? In the early days, we were using Twilio and VoiceOver IP to join calls via a phone line. And that is extremely expensive. A single meeting would have probably cost$10, it would have been really expensive. So we went back and we figured out a better way to capture meetings where why not just have Fireflies attend as a participant over the web, and then really pioneered that, like that was novel engineering.

6:09At the time, we were one of the first if not the first people to be working on that tech at the time, then just started scaling that up, we could join meetings across all of the video conferencing platforms. And that also helped with a lot of our virality, because people would see Fireflies in a meeting, and they would ask, what is this? And when people talk about it, they're like, let me share the notes with you afterwards. And that created a flywheel. So one, we figured out a better way to handle all of our voice calling costs. Two, we didn't have to spend any money on marketing until we got to tens of millions of dollars in revenue.

6:39We didn't spend any money on marketing. Three, I wanted to build a company to be very much self service oriented, meaning until we reached several million in revenue, we didn't really have a sales team or customer success team. It was all done through word of mouth, viral loop, and people could just go in, swipe their credit card and purchase the product. So that was one way we kept our overhead very lean in the early days. Afterwards, we started like scaling and we're doing things at such high volumes. We go to vendors, we partner with OpenAI, Anthropic, and we're doing millions and millions, billions and billions of tokens processing through them.

7:14And we were able to get better rates, negotiate better rates. That was one aspect of it. And then eventually, we started building out our own server compute and capacity, which allowed us to continue to keep the product great, keep the prices low. In our entire journey, we've never raised the prices of our core subscription at Fireflies, despite people using the product so much. And if people back then, if the cost of serving a customer would have probably cost us like$100 to support that customer over a year. And we were charging like 10 bucks. The margins were razor thin at that time. And then eventually we got to a place where we have incredible SaaS level margins.

7:55There were customers where they were costing us$100 a month. And that's an annual subscription. Yeah, we had to figure out lots of ways. And we never wanted to disincentivize usage was like the most important things for us. People love it enough, they care about it, let's give them as much value as possible. When you give someone incredible value at a very affordable rate, they're going to stick around, right? Think about the type of people that go to Costco or buy on Amazon Prime, they're value buyers. We got so much advice from VCs in the early days saying, you need to go heavy enterprise, top down sales, and you need to only target customers in segments where that can pay a lot of money, like insurance or contact centers, or sales organizations.

8:36The problem is every other startup at the time was going after the same ICP, like VP of sales. You're a VP of sales or a VP of customer support, you were getting so many inbound emails and calls. And I said, No, no, no, I want every person in an organization to be able to use Fireflies. And back in 2020, 2021, vertical SaaS selling to a specific ICP just made a whole lot more sense. And that's because if you have salespeople and customer success and account management, all of those sort of functions, you're selling to multiple ICDs. It's very difficult, right? Whereas if you know you're selling to one person inside an organization, it's a top-down sale, it's very replicable.

9:15In our case, because we were horizontal, because we were self-service, because we didn't need salespeople or CSMs in the early days, we could sell to everybody. And anyone that wanted to come in and sign up and use Fireflies could use it, any knowledge worker. And so we really went horizontal against VC's advice at that time. And we really went all on going after a much bigger vision, a much bigger piece of the pie, as you can say. And when you launched the feature to allow someone to invite Fireflies to a call as a user, you obviously had built in distribution by doing that. Did you see literally like a hockey stick takeoff at that point?

9:52Because I imagine having built in virality, if you can get your users to bring your products to other people, it makes a massive difference. So that must have been a real inflection point for the company? And did you literally see that in the numbers? Early on, we had a map where we could see where Fireflies was being used. And then we would see how Fireflies would start spreading in different regions. Remember the first couple months, we saw like a random usage spike in Atlanta, Georgia. And then all these companies in Atlanta, Georgia started using it. Because one of our users in San Francisco had a vendor in Atlanta, Georgia that they would talk to.

10:29So slowly just spread across the entire country. And we'd have like these fireflies light up on the map. And then eventually it went global. We're used by one in three unicorn startups in India. And India became a rapidly growing market for us. And it's growing even faster now as we've introduced 100 new languages. And we support a lot of the native languages. We saw it grow in London, in the UK, before we even had any presence here. Before we even hired anybody overseas, we just started seeing incredible presence here. Brazil was another great place. The Middle East, we're seeing so much usage.

11:05So we became a global company on day one, most companies focus on selling into the US markets, North America markets, and then they try to go into other regions. So because of the viral nature of the product, we had people in 100 plus countries using Fireflies within the first year itself. And now trying to build a global org. The other interesting thing about Fireflies is we're We're fully remote. We're in 20 countries, 50 cities, and we could build an organization that could work around the clock 24 seven. And whether we need customer support or sales or CSM for different regions, we have that built in.

11:39And I think that's one of our competitive advantage. Our services don't stop after 5pm. We have volume and influx of customers having in fact, by the time all of the US meetings end by 6pm, a fresh wave of calls start happening in India, and then Europe. So our system has to be up 24 seven all the time. And maintaining that managing that for the last six, seven years has been very, very taxing. So it's not an easy infrastructure problem to solve. Sometimes the most elegant solutions that end users see have insane amounts of infrastructure and engineering work that goes on behind it. If you're using WhatsApp, think about how many messages they have to process and how much uptime they have, you can have absolutely zero downtime.

12:22Google search at the end of the day is one search box, but it has to be always up. And we have a thing inside Fireflies that I can show up late to a meeting by one or two minutes, but Fireflies has to always show up on time. So that's a tough task as an engineering team. And we've kept our team small, but been able to scale to millions and millions of users. And really, the credit goes to how hard our engineering team works and how we've architected our systems. On the surface, it looks easy to build the note taker. But there's just so much work that goes under to scale it to where you want to get to.

12:53Yeah. And you mentioned that you have made everything self serve from very early on. And you sort of delayed trying to bring in sales team because you wanted to force yourself to make the product very self serve. And maybe that's linked to the type of products that you have that is this kind of like globally spreading thing that you kind of anyone should be able to come and literally get off the shelf and get going straight away. But I know that you're quite passionate about how you actually go to market and the timing of founder sales and then switching over to sales teams. So tell us a little bit more about your thinking when you were growing, obviously, it had built in distribution.

13:29But how did you actually go thoughtfully about growing the company? I think the biggest blessing was the word of mouth and the viral nature of the product. When people see it in a meeting, talk about it, that just naturally spread. And most companies have a challenge around marketing, distribution, lead gen, customer acquisition. They have to do all these sort of things like do performance marketing, SEO, they have to look at different channels. We always said, we're product first, engineering first. So we know that sales and marketing, everyone's going to do the same thing there. So let's be different about how we engineer virality into the product.

14:05And that's what really helped us. This is also one of the reasons behind giving more away for free on Fireflies. Today, we give unlimited recordings, unlimited summaries, unlimited transcripts. Remember what I said, it would have cost us like$100 a month to support a user back then. Now you're getting all of that on the free tier with some limited storage. As you can tell, like my entire focus was on every single person, whether you're a paid user or a free user, you are bringing incredible value to Fireflies just by using Fireflies. And you are the evangelist, you are the ambassador. So we want you to have a delightful experience.

14:40And then you use it enough, you see enough value from it, you eventually will upgrade, you'll eventually hit certain paywalls, or you'll want some configuration or workflows or integrations, where you're naturally going to move up the chain and upgrade on your own. So it's a natural conversion process. And I think that is a lot easier than trying to force salespeople to sell you something that you don't need. The other really important thing is when you're in a new category, like note taking AI note taker, we were one of the first people to use that name and create that category. In fact, we own the domain for AI note taker.

15:14So when we're building that category, the challenge is also a GTM one, because how do you get people to go convince someone to try a completely new category problem? And so we said the best way to convince someone is when you see someone else using it. And when you make it as frictionless as possible for someone to sign up, try it out, because me sitting down and explaining to you the benefits of Fireflies and the AI note taker would not have worked in creating a new category in the early days, people are going to be skeptical. The most important thing was for you to be able to try it, use it and buy it yourself and be like, this is amazing.

15:47And then you go tell your colleague or your friend or your co worker. And that's how it spreads. I think all the best products in the world that have reached some level scale, even if they don't have viral loops built into them, they have fantastic word of mouth. That is the ultimate form of marketing. So chat GPT in the early days did not have any viral loops, you're just chatting one on one. But the product was so magical. It was like an aha moment that you can't contain your excitement. And you're going to go tell every person. Same thing happened with Google, there is no viral loop built into Google, you use search and say this is the best search engine I've ever used.

16:18And then you go tell everybody. Ultimately, if you take care of your users and your users have a delightful experience, a wow experience, then they'll go tell other people. That's the only successful way I know that works when you're building a startup. Maybe you could do a bit of myth busting for us because when you're using Fireflies, you are recording someone, but obviously you're doing it in front of them. You've added Fireflies to the call and it's very visual. So they are aware, but how do you communicate to the market? if you had one message for everyone listening about the sort of compliance or the data protection law around AI note taking, what's the key message that everyone should really understand?

17:01I'm really glad you brought that up. Compliance is extremely important. It's very different in every state, every country, and you should follow your local compliance laws. And we say that in our terms, we make it very explicit. Having the note taker join the meeting and you see the note taker there, at least people are privy to the idea that you are recording, transcribing and having the AI take notes. Fireflies also sends a message once it joins the call and says, hey, I'm taking notes for you. You can also turn on a setting where you can notify people an hour before via email that Fireflies is going to join the call and let them opt out if they want.

17:35So all of that stuff was intentional because we wanted people to have transparency. And no one wants to be secretly recorded. And that's not our purpose. This also was one of the reasons why for the longest time, we did not do a desktop app, because some of the newer age tools that use the desktop apps are secretly recording, like customers, when they use them, the people that they're with don't even know that the call is being recorded. And then you have five, six random call recordings flying out in the wild. And there's no account management, there's no admin controls. So when we built Fireflies, we thought about it from an org centric perspective, privacy perspective, compliance perspective, ownership perspective.

18:14So when you're in a meeting and an admin has control over that, admin can also control who uses Fireflies, who doesn't use Fireflies inside an organization. This is why really serious businesses, when they have to decide between the different note takers, whether it's a desktop app, or whether it's a bot, they pick Fireflies because we really respect the needs of the organization, the compliance, the security. And we have to go through these 1000 questionnaire audit like reviews before they select us. And a lot of times we beat out almost every single vendor, because we built with security and trust in place and compliance in place.

18:51And our enterprise offering that we put out there has some of the most robust admin controls, we have private storage where you can store the data on your own storage buckets, we have a HIPAA compliant product for doctors. So So these things are really important. Another thing I will say is you need to go read the terms of service and data privacy policy of some of these companies. Recently, one of the more well-known companies in the note taker space that's been up and coming who doesn't have a bot, their entire positioning is around silent recording. And again, it has a lot of compliance issues around it.

19:23But if you have that, one of the things that they have in their terms of service, and it just blew up all over Twitter, is that they said that they could be using your data to train their models or train on that customer data. Fireflies, on the other hand, explicitly states that we do not train on your data. We do not train models on your data, you own your data, you want your data deleted, we'll delete it for you. So there's a lot of AI companies that are also using your data, especially needing data, which is super sensitive. And if they're training models or doing whatever, God knows whatever on that, that doesn't create a whole lot of trust.

19:55I'm not just talking about note takers, but even in any other products, just you have to know if the company is using your data for training your balls and stuff. And a lot of people don't make it explicit, they buried that in the terms of service. So that's one of the other myths is just because you don't like having a bot on a meeting doesn't mean that you're automatically compliant when you use a desktop app. So in fact, there's more onus on you now to inform the people that you are recording the call or transcribing it and having notes generated for it. We give people the option, our preference is you use the bot.

20:26If you don't want you can use our desktop app, which is really popular, and it's gaining a lot of traction. It's seamless. You can use our Chrome extension. We have our mobile app for in-person recording and taking notes. You can upload audio files and have them transcribed and summarized within a minute. So there's many, many opportunities for how you want to use Fireflies. We're also doing some hardware partnerships, which we'll like to announce soon. So all of those things are there. But at the end of the day, the compliance thing is ultimately, it's on you as a user to inform the person you're using it.

20:55But also, if you're using something like the bot, the responsibility becomes a little bit easier. So that's why we chose the bot approach more than anything else in the early days. I think it's a really important message for everyone, because everyone is using new AI products, and we're all getting used to these new tools and everything. But only now is the awareness of this kind of compliance and data privacy and stuff actually catching up. And everyone's talking about workflows and agents and everything. And what's so exciting, I think, is that most knowledge work starts with a conversation, whether it's in person or online, it's the genesis of a project.

21:31It starts with a call. It's the genesis of a partnership. So you are right there at that foundation of a relationship or a conversation that leads to real work being done. And note-taking is the tip of the iceberg. So tell us a bit about some of the plans you've got to expand the sort of feature suite so that you can extend Firefly's capability through those partnerships and processes as people do real work? Well, you're absolutely right that most knowledge work starts with a conversation, whether that's over Slack, whether that's over email, whether that's in a meeting. My vision is that while you're having the meeting, the AI is already identifying the tasks and performing that work for you in the background.

22:15So by the time you finish the meeting, you can just say, check, check, check, done, done, done, or delegate it, route it, and then have your work done. We're also bringing Fireflies to not just your meetings anymore, we are bringing it with an email assistant and a Slack assistant to your chat and your emails and bringing all of that context in. We just announced the other day, daily briefs, which is one of my most favorite features where every morning Fireflies goes through all of my past meetings, my past emails and my past Slack messages and threads in the last 24 hours and tells me what needs my attention, where my team is stuck, where I need to take action, and helps me catch up and get a quick roundup of what's happening across the company.

22:57So that's just a great example of how a lot of work originates in conversations. Instead of you having to open seven different tabs and five different inboxes and check different things, you can start your morning off already knowing where you need to put your energy and focus. Our goal is that we can get to those tasks while the conversation is happening. And then Fred, our AI has enough context to be able to help you answer and do those things, whether it's generating a slide deck, creating a memo, creating a follow up email, all of those things can happen with a click of a button, right as the meeting ends or right before the meeting.

23:31Very cool. I think every business should be thinking about how do you get closer to the start of the process because if you can control the start of the process with ai you're going to then be able to capture value throughout and i've seen it in all sorts of things like we're investors in some health tech companies and the diagnostic companies they're there where you find out what's going on and then from there they can sell you supplements physio surgery like the things can go on it's really interesting so i'm really interested in that thinking about how do you get earlier in the flow. And when you're in that meeting, or conversation starter AI space, you're right there.