Krish Ramineni: “It's very important not to just think about the technology first because you want to solve the problem and then figure out what technology can enable it. A lot of times going back to what you can show in a demo with these AI products is really incredible, but what you're not able to do is translate that into a real product at scale.”

Krish Ramineni: “So people always like to say, oh, you guys timed the market perfectly. I always say we actually were two years too early to the market. And most people don't realize that we went through like several years of eating a lot of dirt, eating a lot of shit really, to get to where we are today...”

“…But this credit really goes to my co-founder Sam, where he was bullish that, hey, the cost is going to go down, the accuracy is going to get better and when that happens, everyone's going to rush to it. We need to take that bet. And even if it's a hard problem right now, even if we're going to have to be heads down for two years, if you can just survive and lay the foundation, things will get better. You're like waiting and hoping for a better day, a better future, and it did come, but we were lucky in that sense in terms of our timing was off only by two years, not two decades.”

Krish Ramineni: “I always like to look at three big catalysts for Fireflies. One is the speech technology getting better. That's the enabler—having really great transcriptions. The other catalyst for us was Covid. Everyone started having meetings on Zoom and Google Meet and Teams and everyone went remote. All of a sudden this new technology became front and center and people wanted to use an AI notetaker because they were just going back to back on seven Zoom calls. And then the third enabler was LLMs and then the cost of those LLMs going down. When we started working on this stuff, there was no way we could have used LLMs like GPT4 because it was too expensive. But if you look at in the last 18 months, the cost of intelligence has gone down by 95%.”

“As a founder, you have all of these constraints around how do you build a business where you're balancing between quality and cost and there's like an equation. And that natural equilibrium that you reach eventually lets you get to a place where you make customers really happy and then you figure out how to monetize and how to charge for it.”

Dylan Fox: “Do you feel like it's gotten easier to quantify what that invisible threshold is for where things might not be good enough yet to have product market fit? Now that you guys are a bigger company and more experienced, this might be helpful for other product leaders that are building new AI products… Is it still an art where it's difficult to tell? You have to stay close to users to figure out pockets of… Oh, this is actually still like a gap where we need to find product market fit still, or have you found ways to measure for it more?”

Krish Ramineni: “I mean, there's the standard ways to measure around, like, active usage and NPS and feedback. The real test is just ask the customer if they're willing to pay for it…And you'll know if it's a must have or not. I think that's really the biggest test of everything we do is this good enough for you to want to pay for it? Because I can't give this to you forever because there's a real fixed cost to this…”

Krish Ramineni: “You realize you have to make these trade-offs. Do I go spin up my own in-house models, which we've done experiments around and we have the capabilities to do, but by the time you go through one iteration cycle and get to where you need to for the market, the market has moved along. Like this is an industry where if you just blink, it changes and you will be left behind.”

“So I took that decision that it's best to work with best-in-class providers of LLMs and other models and ASR and all of that and do what we do best, which is serving our customers. And there is a lot for us to do just there. And I don't think of that as being an AI wrapper or being over-reliant on other platforms because our value actually comes from how do we delight and service our users in the fastest way possible. And through that you're building your own moat. You're building incredible workflows and experiences for end users to use. Ultimately that's what makes customers happy, that's what leads to a better defensible moat and that generates more revenue for us.”

Krish Ramineni: “We said, look irregardless of what the outcome is, we're just going to go for 12 months and we're going to build this out and we're going to just incrementally chip away at the problem and make it better. And then, you know, one fine day you start seeing your first person that I never talked to. They go through the application, they swipe their credit card and they pay for the product and they start using it and they say good things about it or they tweet about it. And you're like, if we can get one, we can get 100. If we can get 100, I'm sure we can get to a thousand. And that's all we did. Like, we just started setting mini milestones for ourselves. In this world of AI for entrepreneurs, don't be afraid to go after something that seems technologically impossible or sounds far-fetched because you don't know what the world is going to look like one year from now, let alone five years from now.”