Thought Leaders in Artificial Intelligence: Krish Ramineni, CEO of Fireflies.ai (Part 2)

Sramana Mitra: Are you going after enterprises or SMBs? How do you go to market?

Krish Ramineni: It’s a very blended mix. It’s a very interesting challenge and opportunity with product-led growth. Fireflies can be used by one-person SMBs as well as 10,000-strong organizations. Typically, it starts with that one user inside an organization who shows is to the team. We have this bottom-up adoption. What’s that led to is folks reaching out saying, “I’m seeing a bunch of my teammates using this. I would love to learn more. I want to see how we can deploy it for the team.”

We want every person in the organization to be able to use it. Instead of hundreds of thousands of dollars for a sales tool, Fireflies is easy to get started with. We cover different use cases and spaces whether that’s sales or recruiting – anywhere where you have meetings. We enable customer workflows. For example, if you’re a project manager, you might turn on the Asana integration. If you are in sales, you might use the CRM integration. We take a very broad cut, which is atypical of vertical-specific companies.

Sramana Mitra: Where did you start?

Krish Ramineni: We reached out to a few companies in our network. We would ask them if this service is valuable to them. We had humans do this. That was to validate the idea. It was a very interesting service where we had humans in the loop. We evolved where we realized that this is working. Then we automated it. Today, Fireflies is automated. Almost all of our growth happened organically. It started with a few small startups and then expanded to other startups. We haven’t spent on paid advertising.

Sramana Mitra: Excellent. What is the technology and how accurate is it? We’ve seen this idea many times. In fact, we’ve had companies in our portfolio that were trying to do this. There was an accuracy gap.

Krish Ramineni: The nature of the problem evolves. The reason we had humans in the very early days is because we weren’t sure about the accuracy. The most important thing is that it’s always a tradeoff between cost and quality like most things in business. There comes a point in time where you cross that threshold. We saw the same thing happen for storage back in the day. Then we saw the same thing with computing.

The reason you have these incredible deep learning models today is that the cost to compute is much more feasible than what it was. I definitely believe in Moore’s Law. As the processing and compute get better, you’re going to be able to also train datasets to be more accurate. Over the last three to four years, accuracy has skyrocketed as a factor of overall technology trend.

The other part of it is how you’re able to orient Fireflies for specific types of meetings. Is Fireflies being used in real estate? Is Fireflies being used in tech? How do you detect the custom vocabulary? Getting to 85% can be done with just the basic stuff. Getting that last-mile coverage is where all the hard effort and work happens. Today when you look at where speech is, it’s definitely 90% plus.

A lot of other providers are also getting there. I’m happy for that because I think speech should be democratized. There are certain use cases where you need absolute perfect transcription. Then you are going to hire a person and pay them $60 an hour to transcribe. There are certain use cases where 100% is needed. For most of our customers, 90% plus works.

Sramana Mitra: What is your top use case? Is it project management?

Krish Ramineni: Knowledge management. If you’re in sales and you want a bunch of your sales conversations captured, transcribed, searched and analyzed, you want to them be able to share that with your new sales reps. You can go into Fireflies and leave feedback. That helps with asynchronous training. That doesn’t just have to be sales.

I would do the same thing if I were in recruiting. My recruiter might have an interesting conversation with a candidate. They would share that with me. It would help me get more context before my interview with the candidate. So often, you end up asking the same questions over and over again. It helps bridge the gap there. We’ve also seen inter-team use cases where customer success gets feature requests from a customer.

Then the product team, who usually should be spending more time with customers, is now able to listen to the voice of the customer. It’s way more powerful. What we’re trying to do is help you understand the voice of your customers, the voice of your people, and the voice of your organization. We look at it in terms of those three buckets.

This segment is part 2 in the series : Thought Leaders in Artificial Intelligence: Krish Ramineni, CEO of Fireflies.ai
1 2 3