← The SaaS Podcast - Growing Profitable AI SaaS & AI Agents
The SaaS Podcast - Growing Profitable AI SaaS & AI Agents
458: Read AI: The 8-Figure Playbook for Product-Led Growth - with David Shim
00:56:32
Loading summary
Transcript64 lines
- [00:00]
A
Foreign.
- [00:10]
B
Welcome to another episode of the SaaS podcast. I'm your host Omar Khan and this is a show where I interview proven founders and industry experts who share their stories, strategies and insights to help you build, launch and grow your SaaS business. In this episode I talk to David Shim, the co founder and CEO of Reed AI, a meeting intelligence platform that helps teams capture, analyze and act on insights from their meetings. Before founding Read AI, David sold his first startup for $200 million and later served as the CEO of Foursquare. But when he started this company in 2021, he wasn't chasing another big exit. He just wanted to fix a problem. He felt every day too many meetings that went nowhere. During the pandemic, he found himself stuck in back to back video calls that weren't a good use of his time. That's when the idea hit him. What if you could measure engagement and distraction during meetings, but read AI's first product missed the mark? It showed engagement analytics but didn't tell users what to do with them. Their first month retention was just 5%. David realized that he'd built a data dashboard and instead of a decision making tool, then OpenAI launched ChatGPT and over a weekend his team built a meeting summarizer. It worked, but David saw the trap. If they could build it that fast, so could everyone else. He decided Reid wouldn't compete on summaries. It had to go deeper. So they focused on what transcripts miss, tone, emotion, reactions and engagement. They basically built technology that told the story of the meeting, not just the notes. The first month retention climbed from 5% to 30% and then 40% and eventually to 81% as they added features like auto highlights and head nod detection. But instead of chasing revenue, David focused on day one, ROI and virality. Today, Read AI adds about a million new accounts a month, generates eight figures in ARR, and does it almost with no marketing spend. In this episode, you'll learn why his first product failed and how he fixed retention by turning data into actionable insights. What shift helped Read AI stand out from the flood of AI summarizes and reach 81% first month retention how he used product virality instead of marketing to drive over a million monthly signups. We talk about how he first built a product people couldn't stop using before focusing on monetization and how David differentiated Read AI to successfully compete against big players like Google, Microsoft and Zoom. So I hope you enjoy it. Compliance isn't optional anymore, but it also doesn't have to slow you down if you've ever wrestled with SOC2, ISO 27001 or, or GDPR, you know how painful and time consuming it can be. Sprinto automates the entire compliance process, from evidence collection to vendor risk across more than 300 tools. You stay audit ready, secure and focused on building your product. Trusted by over 3,000 companies worldwide, Sprinto helps SaaS teams breeze through audits and keep customer trust strong. Book your demo@sprinto.com that's sprinto.com David, welcome to the show.
- [03:44]
A
Excited to be here. Thanks for having me.
- [03:46]
B
My pleasure. Do you have a favorite quote, something that inspires or motivates you that you can share with us?
- [03:51]
A
This is a version of a Muhammad Ali quote, but it's really going in and saying struggle now and live the rest of your life as a champion. And I think that helped out a lot when I was a kid growing up to where. Now if you're a startup and if you're a SaaS startup, there is a lot of struggle. But if you can actually get past that and really kind of find that product market fit, you live your rest of your life as a champion.
- [04:13]
B
Yeah, I love that. So tell us about read AI. What does the product do? Who's it for and what's the main problem you're helping to solve?
- [04:21]
A
Yeah, so read AI is for everyone. And the way that we think about it is it's a system of record, of intelligence. So it's all the data that you've generated through your emails, your messages, your meetings, Asana, Jira, Confluence, Gmail, Outlook, all that data is in disparate data plate sources right now where the kind of siloed, we bring all those together and then we go in and we say how do we make that actionable? And so the most obvious actionable actionability is with meetings. We're able to record meetings, we're able to give you summaries, action items, follow ups. But then you think about it, it's like what happens after that? Well, if you only record the meeting, you don't know that I actually followed up with the IO that I followed up with a RFP response. And so you can check the box that meeting that it got followed up with. But then after that email goes out, maybe someone internally asks you a question like hey, what's the status on this rfp? You know what's going on? You've got that data now that you have to go in and call them and get that information and you've got that recorded like, so it's about Bringing all those data sets together and then being able to be that system of record to drive productivity, to drive intelligence, to drive actions at the end of the day. And I think when you think about AI, right now everyone's talking about agents, but if you think about agents from an AI perspective, agents without a specific purpose are unemployed. And so what you want is really agents that have a purpose be taking meeting notes, updating Salesforce, updating HubSpot, going in and reminding you to send an email, reminding me that you should follow up three days later and you don't have to ask for it. That's the biggest thing too. Like, I think this chat component is great right now, but Ask Jeeves only last so long and people want the content delivered to them almost like a TikTok where based on what you do during the day, during your work, you want that content delivered to you. And that's really what Reed is about. It's about AI in the background, where it's delivering based on what you need and predicting what you need.
- [06:12]
B
Yeah, I mean, that's the struggle that I'm seeing more and more myself is that like I'm obsessed with this idea of just like AI and automation and everything that I can do with this podcast, for example, to get better at just the workflow and everything pushes me back to a chat. And it's like, I don't want to go there. Right. I want that stuff to come to me. So I remember to do these things. And there are ways around that. But it just doesn't feel very natural or intuitive at the moment.
- [06:40]
A
No, a hundred percent. Because right now, if you want to say what is the most natural way to consume content, to make decisions, if you're going to say globally, not just for SaaS, not just for enterprise, but across the Globe, it's either TikTok, where the content is a feed, or Instagram where you get reels and it learns you never do a search generally unless you're like, that content's interesting. I want to find more from that person. And it serves the content based on how long you interacted. Did you like it? Did you comment on it? And it picks those things up in the background. So I think that is where the future of AI is going. And then the decision making process is Tinder. If you think about how many decisions are made around Tinder, it's like, swipe right, swipe left, apply it to hinge, apply it to any of these other applications. But that is your life. Who am I going to date? Who's Going to be my significant other. You're just swiping back and forth. So I think when you combine those two things together, a year or two from now, SaaS is going to look a lot like decision making that does Tinder and TikTok combined.
- [07:30]
B
I love that you just summarize the future of these agents in like two apps. Give us a sense of the size of the business. Where are you today in terms of revenue, customers, size of team?
- [07:43]
A
Yeah, so we're adding about a million net new accounts every single month. So we're on a run rate of about 12 million accounts created for 20, 25, 26 depending on the month that you start it. From a revenue perspective, we're at eight figures ARR, we're doubling tripling year over year. So overall kind of, we're kind of in that compressed hyper growth period that you hear about a lot of SaaS companies that also layer in AI that you see in the market. I think what makes us a little bit different from kind of everybody else would be that we spent almost zero in marketing today. Like we've done some tests here and there where we spend like five grand, ten grand, but that 12 million run rate on a signup basis is based on zero in ad spend for the most part. So it's just this organic PLG motion that the product is driving enough value that people are able to test it, they're able to see the ROI that they're like, okay, this makes sense, I'm gonna go buy it.
- [08:35]
B
I love that. And we're gonna talk a lot about that and what size of team.
- [08:41]
A
We're a little bit under 100. We wanna get to about 115, 120 by the end of the year. I think we'll see if we get to that number, but that's the goal here.
- [08:49]
B
Okay, great. So before we get into the story of Read AI and how you started that business, you have a super interesting background. You've worked at all of these startups and companies that a lot of people know very well. And we can't talk about all of them right now, but in 2011 you launched placed and you worked on that business for about, I think nine or 10 years. During that time it got sold to Snapchat and then eventually to Foursquare. Just tell us a little bit about that product. What was it about? How did you come up with the idea?
- [09:34]
A
Yeah, that idea is really around the location space. And so if you think about it, iPhone One had come out in 22, was about to get released or released that year. Android was just getting introduced. And I had this idea in my head that if the phone is a cookie, the places where you go are websites. So like a Starbucks is a website visit, but it's in the physical world. So if you combine those two things of like the person is with their phone, they're going to these places that should be valuable. I didn't necessarily have like a business use case other than the analytics are interesting. So I kind of kept on going and I kept on sending developers out to these conferences to say, hey, when is it going to be possible to run an app in background to get location data persistently? And their first conference they went to, not possible. Second conference, not possible, third conference, like, well, there's a thing called Android. They're going to let you do that and iOS kind of lets you do that now. And then I said, okay, that's what I wanted to see happen in the marketplace. So I decided to then leave my job at a company called Great Company called Quantcast. And I said, like, all right, I'm gonna build a team. I'm gonna go in and try to figure out how do we get this location data persistently in background. And you might say, like, that's kind of creepy, David. You're getting location data back, especially back in 2011. But we really solved for that was like I just gave people money for their data. So it wasn't like this sneaky thing of like, hey, let me try to get your data without telling you. It was like, hey, I'll give you anywhere from a dollar to $5 per month that you can buy an app. You can buy five apps if you want to, every single month. As long as you let me measure where you go. And if you confirm the places where you've been, we'll give you a little bit more. And so what we were able to do is build this huge training data set where people were checking into locations. We had the raw location data on a persistent basis. And then we built these models to identify store visits. And so we did that in 2011. We were Wall Street Journal, New York Times, et cetera, did not have product market fit, not from the sense that people were interested in it, but people weren't buying it. Like our first client was Best Buy. They bought it to go in and say, hey, on Black Friday, we have all of our executives go out to these individual stores and count how many people are at our stores as well as our competitors stores. And then we survey them and say, where are you going Next to understand, like, how people ebb and flow. That year where they started using Reed, nobody went out in the field. So we were able to kind of automate that process for them. But there aren't a bunch of companies that have that use case. So we ran into an issue of like, hey, this is great intelligence, but how do you use it? And then ultimately what happened was I didn't want to build an ad business and we actually didn't build an ad selling business, but we built an ad attribution business to say, Spotify. If you serve an ad for McDonald's, wouldn't it be valuable to tell McDonald's this is how many people went into the store after they heard the ad? And so that really snowballed into this business that we became the default attribution solution for location across billion dollars, billions of dollars in media spend that Snapchat came along and said, like, hey, we'd love to buy you. We were trying to build the double click for the physical world and we think you'd be a great addition. So that's how the acquisition happened.
- [12:32]
B
I don't know if you can talk about how much they acquired the business.
- [12:35]
A
For, but we're far enough away now that I can be a little bit more liberal. So it's for about $200 million, depending on the day and the stock price. So they acquired us for 200 million. And then from there we were in their. We were in their org as a standalone company, so it was rare they didn't roll us up. But they said, we value the independence that you have. We want you to continue to work with all these pub and advertisers without having to buy Snap Advertising. So they said like, we'll keep you as a separate entity. That actually ended up being a good thing. During that time, Snap went down to from like 20s to $6 a share. Not because of us. We were growing fully profitable. We had been profitable for five years prior to the acquisition. So when they acquired us, they brought a profitable business and we were contributing to the bottom line. But in 2000, was it 19? They said, hey, a lot of people are actually pinging us to say, we want to buy this business from you. And we're willing to pay pretty much the same amount, if not more for what you actually have. Because this is the market leader. They're still growing. So we spun out into Foursquare, was the person that ultimately acquired us. And then I was at Foursquare for about two months. And then they came along and said like, hey, we'd like you to be CEO. And I said, okay. Like, it wasn't expected. I was. I was more planning on putting it into a good spot, but, you know, that opportunity was one I couldn't pass up. So I became CEO Foursquare in 2019.
- [13:52]
B
Cool. And then I think maybe a year or so later, you got the. You had the bug bite you again to go out and build another startup. So, like, that was like, read AI. So where did that idea come from?
- [14:07]
A
Yeah, so what happened was about a little under two years in that role, a lot of meetings, a lot of interactions, and about six months into six months before I left, I kind of gave the board the heads up. Hey, I'm departing. Heads up. And then I started thinking of different ideas, a lot of them, nothing actually came to fruition until I left Foursquare. And what was interesting was like, once you leave a company and you have a couple exits, a lot of people either want your money to invest in, they want to sell you stuff like an advisor, they want you to be an advisor, they want you to be an investor. There's all these conversations that I had. And this was peak Covid. So this is like 2021. And so everyone think was on zoom video conferencing calls. And what I noticed on some of these calls was within about a minute or two, I realized I should not be on this call. This is not a good use of. Is a bad use of time. And I want to leave. But I kept my camera on. So now if I leave, it's going to look weird. So what I started to do is surf the web, but there wasn't a lot of stuff going on because it's Covid. So there was a lot of activity. But I was on ESPN and I noticed someone had glasses on. And I kind of saw the glasses and the colors look kind of similar to what was on my screen. So I double clicked on their zoom little image and I could see the words and I can see the colors and I could see the picture. And it was espn.com and so for me, that was kind of the magic moment where it said, like, okay, we're not spending our time wisely on this call. There's a lot of senior people on this call, and I can measure that you're distracted because I'm also distracted now. Is there something to build from there?
- [15:31]
B
Okay, so you were seeing the reflection on this person's glasses, and they were.
- [15:35]
A
Browsing espn and I was on espn. I could confirm it.
- [15:40]
B
Love it. Okay, great. Again, it Wasn't like some huge flash of lightning. It was just some problem that you saw there was a potential opportunity. How did you get started? What did you decide to do to either go and validate the idea or build a prototype?
- [16:05]
A
Yeah, I think one of the first things I did that a lot of early founders don't do is they think their idea is really brilliant, unique. No one's ever done done it. And so over time I realized, like, there are companies working on different projects that might be very much aligned or failed. Aligned, failed and successful. Like all these variations. And for me, I did a bunch of searches and I said, like, are there models that exist today to go in and actually take a multimodal approach? Multimodal means video and audio and process both of those to understand things like sentiment and engagement. And as I started to kind of go down and look at the different papers, there's a lot on transcription. Like, if you think about Alexa has been around for a long time. Before AI models, before video conferencing was widely spread, like there was natural language processing. So that existed, transcriptions existed. So that was like, okay, a lot of people are doing that. But when it came to video, the only things that were out there from a model perspective was one taking security video and kind of trying to identify when people are walking in different areas to go identify shoplifting. There were things where it's like commercials. Where you watch a commercial, it would track your eyes to say, are you following the commercial? Are you looking around and you're disengaged. But none of those were like production models. None of those were real time models when you want in a meeting. And so I said, okay, this is kind of interesting. There's not any model designed for this world now where video conferencing is 100% of meetings during COVID And so I was like, that's good, that's interesting. There's. There's no analytics that I've seen in market around there. I did a bunch of research, There's a bunch of transcription, but there's no analytics. So I said, okay, big market, hundreds of millions of people every single day on this platform, but there's no analytics. So then I called someone when I first became CEO of Foursquare, Eric Yuan, and Zoom reached out and said, hey, congrats. I just want to say it's awesome opportunity. I know Foursquare, just introducing, if I could be a resource, let me know left Foursquare. And I said, I was like, who can I ask that would know this better? Than anybody else. Is this one, an interesting product? And then two, how hard is it? And three, are they working on it? Are they actually building a competitive product? So I hit them up and we jumped on a call and I asked him, like, hey, this is what I'm working on. Would love to get your thoughts on. Is this valuable? And two, are you working on this as well? And great part about him was like, he's super transparent. He was like, we thought about it pre Covid, but we didn't actually build it out because Covid hit and we had this bigger opportunity to go after to be kind of the video. Video conferencing platform for the world. Great, great, great answer. So I was like, okay, you're not going after it, but it was a market that you thought was a big opportunity. And then two, I said, hey, are you going after this? He, no, you should go after this market. It's a big opportunity. So I think those two things combined. Kind of went in and said, like, all right, now I've gotten validation that this is something that people want that other larger entities were going to do. Had it not been for this really kind of black swan moment with COVID Then I said, okay, now how do I go in and tackle this? And so that was kind of the validation that was able to get to go in and say, this is something I want to go into.
- [19:00]
B
Right. And did you raise money? Did you raise a seed round, like, right away?
- [19:05]
A
Once it got the validation, then I went out and said, like, okay, I'm going to go after this. Let's go raise a round. We raised a $10 million round in 2021 that was led by Madrona Venture Group, and they had invested in my previous company, placed where we had a pretty good exit. And so from that standpoint, kind of known entity, they saw the market and they were kind of like, okay, this is early. This is a big market to go after.
- [19:27]
B
Cool. Okay. So you ended up spending most of that money on the product. It wasn't on marketing. So first, let's talk about the first version of that product. Like what you described read AI to be today. It wasn't that in the early days, right. It was kind of more like another note taker. When you started out, it was more.
- [19:52]
A
Like a car dashboard, which is kind of weird when you think about meetings. But we have the analytics for sentiment and engagement. We have a couple patents filed for that. So we've got these kind of foundational models, not the right word, but these core models that we've Built out to go in and measure sentiment, engagement. And what we found was we were really accurately. People were like, this is really good. And people started to come to us when we launched it on the Zoom app store. So it was this app that's on the sidebar of Zoom and we launched it and people were like, oh, this is actually really accurate. I actually think this is really cool that technology can do this. And the use cases were things like salespeople on a call. If a call goes bad, like, I can actually see that my audience is disengaged. Because a 16 inch laptop monitor does not show me all the people on the call to say, are they happy, are they sad, are they engaged? Or are they disengaged? For teachers that are teaching college classes, they were doing all their classes remote. They want to know when they should stop and ask the question, like, does anyone have any questions? Or hey, let's talk about this in breakout groups. Because I've lost their attention. But the problem was the AI wasn't fully there to say what you should do next. So we were identifying that a problem existed or an opportunity existed by measuring sentiment and engagement. But people were starting to get freaked out. They're like, I believe you. It's going from a 75 to 74 to a 68 to a 62. It's getting worse. You need to tell me what to do because I'm getting like lost in the numbers now where it's like, what's going on? In the same way, like if you were driving a car and all of a sudden you're just like looking around, you're driving normally, you're like, I know how to drive. But then you've got all these dials, like accelerometer, gyroscope, you know, speed, speedometer, odometer, all these things. Like, what do I look at? What do I do? From an action standpoint, we didn't have that. So we, we looked at the app and what we found was we were getting like 5% retention after 30 days for a free product. And so on the Zoom App Store, they launched it. We were, they had an app store. We launched it as one of the inaugural apps, so one of the first 12 apps. So one thing was like that distribution was incredible that they got us in front of a lot of folks where we were able to kind of, without spending a dollar, get in front of millions of users that would try out the application. And for us, the question then became, do we charge for it right up front or do we just make It a free product. And because we had raised some of the money from Madrona, our approach was to actually go in and say, let's just make it free because we have to see if this product is sticky. And back then the tooling wasn't there as it is today with all these AI like lovable replit, et cetera, to just kind of spin something up. It was a lot of work. And so but we decided we had enough funding to go, let's make it free. Let's see what the traction looks like. We saw a lot of users create accounts, but the retention wasn't there at 5%. So it's kind of like we would make iterations, we would say like, hey, what if we actually put transcripts in there now, not just the scores, and that got it a little bit better. What if we did video got it a little bit better, but it never really got above that 10% threshold when we were working on the core product. And then the iteration was in 2022. OpenAI came out and said like, hey, we've got this thing, you can try it out. We tried it out over a weekend and we're like, oh, this is actually really cool. You could load up a transcript, it'll summarize it for you. But then I asked my team the question, like if we can do this in a weekend, every single person in the world could do this by just taking the transcript and putting a prompt in and summarizing it. So what value are we actually producing? And so it's kind of saying front running, like, hey, this is going to be really cool as the first mover, but long term you're going to get a ton of competitors from a meeting note taking side. And that's true today where if you go to Product hunt, there's probably like two to three that launch a week around meeting note taking. And so I said what can we do that's different? And what was the differentiation was let's take the scores that we generate during the call, how people reacted to the words that were said and generate a narration layer. So think about transcripts, tell you who said what, but it doesn't tell you how people reacted. It doesn't say how were the words said. If you ask me a really tough question and I stop for a minute and I started talking really fast at 225 words per minute. The transcript would just say, David said these things and that was a great answer because it looks really good versus like oh, his rate of speed went up, his tone changed, he Looked kind of nervous when he said it. People started to get a little bit annoyed based on things they were saying. Now all of a sudden, a transcript that you put in ChatGPT say, great, you should buy the product now became, David said some really interesting things, but the audience really didn't buy into it. How do I actually solve against that? So that's where it really changed things. And we saw this really huge uptick in adoption from there.
- [24:17]
B
What did I do to the first month retention?
- [24:20]
A
So first month retention went from 5 to right under 10 till we were at 30. And then we introduced video clips. So because we can see the reactions of people on a call, we were like, hey, what if we actually took it like almost like a movie where a movie trailer where we can go in and say, find the scenes that people get shocked at, either like super happy or super sad or engagement goes up and then look at the phrases before, like that one or two sentences before, and automatically chop it up. And so because we have the reaction layer, we could chop up this video and we made a 30 second like commercials for a meeting. So if you want to see, should I actually go to this meeting or review the notes, you can just play the video or show me the ESPN highlights. Show me the top three minutes of a two hour long call. We would be able to automatically cut that up. And so that drove retention from 30s to 40s. And then we went from 50s where we were able to go and highlight different points in the transcription that was high engagement. We were able to get better action items. We were able to say when someone nods their head to say like, hey, are you going to deliver that? We can actually assign it to that person. Because we have this multimodal approach. We were able to build models to go in and say when you look away, like here to the top corner, when you're not talking, a lot of people do that. That's because they're actually looking at their second screen and they go back to this camera. So if you go to fixed point, can you actually say they're still engaged as they keep their head here, but if they scan the room, they're not engaged anymore. So all these things we started to stack together where now today we're at about an 81% retention after 30 days. So if you try it in a meeting, 81% of people are actually using it 30 days later.
- [25:52]
B
Wow. And I think it's like that puts you in like the top 10% of products out there.
- [25:57]
A
Yeah, I think Andreessen Horowitz has us as top 50 AI application across the globe. I think we are in that group. I would say we're probably closer to top 25. And then from an adoption standpoint, we continue to see adoption where US is our core market, but we're global. So we have anywhere from 1 to 2% of Colombia's population. Would have never expected that, but 1 to 2% of Colombia's Population is using Reed. South Africa, very similar. And what was really interesting for me was this AI kind of adoption curve is not just enterprises like you would have expected. It's actually broad based consumer and it's across the globe where people are like if you're gonna solve a problem for me. And I think this is where it's a little bit different, where it's not a salesperson pitching something to you, but it's going in and saying like I saw this in a meeting, it's got great notes. Someone else forward me this summary where they pulled all these things together and it was really cool. I'm gonna go create an account. They create an account and they try it out and within like an hour or two they're like, okay, this has roi. I'm going to sign up for the service either as a paid customer or a free customer. And I think that's all done in a day. It's not this long drawn out process anymore.
- [27:07]
B
A quick break to thank our sponsor, Sprinto. Most companies treat compliance as a side project, but without it, enterprise deals stall. Sprinto turns compliance into a competitive advantage. Its automation platform unifies audits, risks and vendors. So you're always audit ready. With 300 plus integrations and AI driven monitoring, Sprinto keeps compliance running quietly in the background. Join over 3000 SaaS teams already using Sprinto. Learn more and book a demo@sprinto.com that's sprinto.com how did you. I mean, you know, it was still the early days of like AI and sort of using that to build products. And so there's a heavy attacks every time you decide we want to build this feature, do this. And in many ways a lot of founders will be able to love to be able to do that where they've got a list of things that they think are going to make the product stickier. But it's an investment, it's a trade off. It sounds like you were pretty rigorous in the way that you were experimenting and identifying these things. Was it as sort of well thought out as it sort of sounds?
- [28:23]
A
I think building of the models, that multimodal model to build the narration layer, that was very thought out, like we've issued patents against it, like, that was the core vision, the actual application of that information. We thought it would be a great standalone and it had some use cases, but really the value proposition is when you connect it to something else. And part of it was honestly timing as well as luck to go and say these large language models came out to summarize things. And there's a couple of different options. You can use OpenAI, Anthropic, et cetera, but they're able to summarize it with different prompts. And then if we've got the secret sauce that nobody else has, which is around this multimodal approach to interactions and you combine the two things together, that actually was great timing because had the LLMs not come out and when they did, like, we might be doing something very different, but it just allowed us to say this macro trend of being able to summarize and condense all this content, plus this unique layer of narration that we're able to put in, really differentiated us. So I'd like to say it was all this master plan that I had, but honestly, it was really more us seeing what was in the market and also us not going in and saying we need to build our own LLM. Like, if you go back to when OpenAI first came out, everyone else, there's a lot of startups that say, oh, we're building our own language model, we're doing X, Y and Z. And those companies are all gone. There's probably like four or five of them of any scale that are left now, but there are probably close to 100 plus where they're like, hey, we've got this, let me open this up. We were working on this internally, but those are all gone now. Now you've got a kind of a set number of leaders.
- [29:50]
B
Yeah. So when you talked about, like, adding features to the product that help drive retention, 30%, 40%. Was it just looking at, like, user behavior and what was going on in the product, or were you going out and talking to. To users as well? How are you getting those data points?
- [30:10]
A
I think there's a balance that you have to take between talking to customers and then looking at what they do. And I think we're going to swing the pendulum, swing to just looking at what they do more as we get into this AI age, because the data is there, you can see how they're using it, where they drop off. And you've got teams that look at it today. But as you start to Identify those patterns. It becomes clearer and clearer. Where's the drop off? Where? Sign up process? Where hey, they go see the meeting report? Are they seeing the subsequent meeting report? Where do they stall out on the page? All of those values are there. And we actually kind of. It was a hard thing to do, but we said let's not ask the customer because this market doesn't exist yet. So to ask the customer what they want is going to build. It's going to be building this comer Simpson car where if you saw the episode, Homer Simpson found his long lost brother. He's a, he's an executive auto company and they worked together to build this car. And Homer added all these features that were great. But when it all got done, it looked like a monstrosity. Nobody bought it. And so I don't want to do that. We didn't want to do that because AI is incredible in theory. You can do anything with AI. But we said let's not do that. Let's solve real problems, let's stick to our core and then let's go in and see how people are using it. Where are they dropping off, what don't they like about it based on the actions that they take versus what they would say from an opinion standpoint.
- [31:25]
B
So as you mentioned earlier, you raised 10 million when you got started, but then you also decided that you weren't going to spend any of that money on marketing. And all this growth that we have talked about has really come from what you and I were talking earlier. Like the $0 sort of marketing playbook all about PLG creating these viral loops and even sort of your whole land and expand strategy you were doing without any salespeople. So we'll go into the sort of details, but why did you decide to not spend money on marketing?
- [32:07]
A
I think you could always spend money on marketing. That's a note that you put a pixel on, you see what it converts at, you test the ad creative, et cetera. But where we really wanted to differentiate was building a product that people would find value in within the first day. And that was kind of like that goal post that we went to is like how do we show as immediate value as possible? How do we solve the cold start problem? Problem where I need you to do 12 things to actually show you any value. And so for us it was really going into the meeting, sending the report as fast as you could, but also making it comprehensive, not going in and saying like hey, you have to come to our website. In the old days what happened was a Lot of these meeting note takers would go in and say, you have to come to our website to see the report. We're not going to make it easily shareable because we want everyone to buy a subscription. We went out and said, like, dude, meeting notes should be for everybody. Like, we should make it easy to distribute this. Because if it's easy to distribute, more people get value against that. And there was like pushback. Well, what if people don't want the notes? Well, everyone was on the call. Like, everyone was on the call. So they all could have taken their own notes. So it's not like this controversial thing. It's like, I can show every person on this call value that Reid was there because I've got the notes and now they can compare it against their own notes and say, what's more valuable? So I would say that immediate ROI was the big thing, and then that PLG motion came from that immediate roi. Because when people see that in action, they're like, this is great. And I'd like to think we're pulling in from some of Slack's early playbook where they said, hey, it's really multiplayer. It's not about single player. It's not about broadcasting things out and then hoping people respond. But it's communicating two ways. It's having that conversation that goes back and forth that you want to add another person because you're like, oh, it sucks. Like, Carrie's not on this call. Let me add Carrie real quick. And Carrie goes in and she sets up an account and now she's able to do access Slack. And I think that level of kind of value and setting it up where it's like, was it 5 free accounts or 10? 10 free people in a network is free on Slack. And then they start to charge for it. And that's where we were a little bit lucky with our track record from our co founders, that the investors gave us a little bit of leeway. They saw that we were trying to build something bigger than just a meeting note taking app.
- [34:14]
B
So give me maybe one or two examples of like how you were building these sort of viral loops into the product. You had basically a free or it was a free product at the time. It wasn't even freemium, right?
- [34:30]
A
Yeah, totally free. You could upgrade. So today the product is totally free if anybody wants to use it. Five meeting reports a month and a lot covers like 80% of the population. Because I think for this podcast, people probably have meetings every single day, multiple meetings a day. But the general population, 5 is pretty close to, like, the max that you would want to kind of have notes for. And so we're like, hey, that's fine. We want to get adoption, because eventually, longer term, we'll have additional features that you might want to enable that you would upgrade for. So that was really important from a freemium standpoint to give people that opportunity to try it out. And what we found, too, this is a metric that was really interesting, was the longer you use the free product, the more likely you were to convert down the road. So you'll hit a point that you get to 10 meeting reports a month. So you've got to be able to take that early hit of that cost per acquisition to say, like, hey, I need to pay for this free service. But at a certain point, we're like, yes, it's great. They've got a year's worth of my meetings. They're storing it for free. I want to query it, but I can't query it unless I upgrade my plan. So, yeah, I'm going to go in and upgrade it, and I'm going to create that loop. So I think it's the loop is the data mode. It's that people are agreeing to go in and use the free product, and they get to a certain point where this is our job. It has to. That data has to be so valuable that you want to upgrade to that next version to get access to that data without having to hunt and peck against different things. And our price point, we chose in the age of AI and you're seeing this more and more. It's not about price points where it's $500 per license, per user, per month. Ours was like 15 to $40 per user per month. So it was kind of an easy thing to bite and is also giving people the ability to sign up on a monthly basis. So it really wasn't that big of a deal for people to spend, hey, 15 bucks. Reid will go and give me all this additional incremental value. And we said, if you want to cancel it, totally fine. Like, so we've got people, like, in developing countries where we haven't localized pricing, we see people, a lot of people gaming it. So they will go in and they'll create an account one month and they'll cancel. Then they'll come back the next month and create an account because they want that service, that those tooling. And we're like, that's fine, because that's actual learnings for us when we go into that market more aggressively and we try to figure out A price point is it usage based pricing is just a lower, low overall price point. But it was understanding that two would be from a kind of. A little bit more kind of. When we started to get product, market fit was multiplayer. So teams want to share data with one another. Like if I may, I'll give you a. We've got a Mag 7 client that we have that has thousand plus licenses. And so they went in and said, hey, we've got 200 product managers. Those 200 product managers have five calls with customers every single week. So just one a day. And right now across the globe people keep that information siloed like it's based on their specific project that they have that they'll share with their team. But someone in Tokyo is not going to share with New York, is not going to share with Mexico City. Not because they don't want it, but they don't. It's too big of an org. They don't know who to share with. It's not available. Well now what they've started to do is actually create these folders where they say, hey, every customer interview goes in these folders. We've used Search Copilot on the read side to actually ask questions like, hey, what are the 10 most common requests this month? Hey, what are the 10 most common requests that haven't been asked in the previous six months? Hey, we've actually connected their CRM system so they can go in and say like, who's the person that actually has canceled their service and asked for these features? So now you're able to narrow it down. But that multiplayer aspect, you have to show the value. But as you start to stack each additional person, the AI gets smarter and that's really that you're probably going to hear this more. But that storage of intelligence, that's where that comes into play is with multiplayer. It's not with single player.
- [38:04]
B
Yeah, I love that. One of the challenges I think a lot of founders face when they're building off a basically on the back of another product like, like Zoom or teams is waking up every day with this fear that that company is going to go and build the same thing into their product. And usually what happens is either they'll try to kind of go too broad too quickly and have a bunch of issues there, or they just do leave things until too late, until it's basically sort of a commoditization thing and they're in real trouble. I think you navigated that really well. So just tell us about that. I think this is a question Right. Which I'm sure you came up in the early days. Like, if Microsoft and Google and Zoom build in these features that read AI has, what's the point of having read AI? So how did you navigate that?
- [39:09]
A
Yeah, I think the most easiest answer when it comes to meeting notes would be over 60% of our users use more than one platform on a regular cadence. So you've got your Microsoft Notes, you've got your Google Notes, you've got your Zoom Notes. If those notes can't talk with each other, which they can't today, if you use those platforms, you can't actually get the knowledge that you're talking about with multiplayer. You can't go in and say, this customer wanted this, but this customer didn't want it, or, hey, I've got 50 customers that want this, but if I silo it, it doesn't look that important. So being able to bridge that information is key. And especially from a sales perspective, you're going to use whatever video conferencing platform the client wants. So if the client says, we're a team shop, we don't do anything else, you're going to use teams. So if you're a Zoom shop, you're going to lose out on all that data. So for us, inherently our user base uses whatever platform is available and you want that almost like the open Internet, you want that ability to actually take that knowledge and store it, regardless of what platform you use that on. So I think that was one differentiator. I think two is we took the approach of they're educating the entire market. So AI did not exist in anyone's vocabulary outside of sci fi fiction in 2021 or 22. Like, people were like, OpenAI didn't exist. No one knew that you were able to do all these really incredible things that we are doing today, but no one knew that existed. But thanks to Microsoft, to go out and say, like, hey, we've got a partnership with OpenAI, we're doing copilot, we're making it available to everybody. That push made this mainstream very quickly. Then Google came along and then Zoom introduced Zoom Companion AI. So they educated the broader market. And that's as a startup to say, this is a nascent space to educate every single person in the world what AI is. That's a huge value proposition. So we kind of rode that wave and we took the approach of, yeah, Google, to take your job away or your company away. Microsoft, Google, that's always been the case. IBM in the old days, even now it's OpenAI or anthropic. What if they release a feature in there? But if you look at it historically, there has been an entire ecosystem said those are the platforms. But you look at Microsoft and you look at Excel, you could say Excel, nothing else should have existed. But you've got Google sheets. But then you go one step further and you go and say there's hundreds of billions of dollars of value on databricks. Snowflake, smartsheet, tableau. If you give list qualtrics, the list goes on where these companies. Because Excel only did one thing broadly, it didn't solve very specific problems. And so I think if you believe that and you just stay focused in on the opportunity of what problems are you solving, that's where you kind of go in and say like, yeah, they're not working on that, or they rolled it out, but it's six tabs in. That's not a core feature. They're just checking the box on a grid to say, we have that now. And the last example I'll give is since Microsoft Copilot launched on Microsoft Teams, we've seen a 20x increase in the amount of meetings that we measure. So it's because they've grown the ecosystem. We've also grown alongside it.
- [42:08]
B
One more thing I want to talk about before we wrap up. We sort of got a few minutes left. Let's talk about your sort of land and expand strategy. So you had people who are coming in and many of them working at Fortune 500 companies and they sign up for a free account. They use read AI and then maybe people in the department are using it. And at some point there's a discussion that happens about having some sort of, you know, licensing agreement. Typically that's where a salesperson would step in and handle all of that. But you did that without any salespeople. Can you like just explain how you're doing that?
- [42:47]
A
So we first three years we had no salespeople. We started to build out a BDR team that handles the inbounds that come in. So we've only just started that function and they're doing a great job. But where we really focus around was like self service. How do we let people triage their own issues? How do we make it easy to set up account to purchase a license? And what would typically happen is people would go in and set up one account, then multiple people would set up an account and then the IT team would either go in and say, hey, we need to set up a corporate account, or they would reach out to us. And say like, hey, what is this? How do we get this organized within an enterprise scenario? And what we did there was just kind of like, we just supported it. And this is kind of rare, it's very rare to see, but it's. They want. They saw that this was being used, they went into us and they said, hey, I need to control this. We gave them the controls to do it on a self service basis. So we just said like, we wouldn't even jump on calls. In a lot of cases we'd be like, hey, if you want to create an account, click on this link. And if you get to X number of like 100 plus licenses, you automatically get this discount. Plus we've got SAML, we've got SoC2, all that good stuff. And they're kind of like, okay, no one's going to call me. It's like, no, if you. This is kind of, you got to be confident in your product. But we are confident with 80% retention rate where it's like, no, this will save you time, this will save your team time. Where they're getting demanded from the IT employees are demanding from the IT department to utilize it. So it wasn't that hard of a sell. I would say. The things that are kind of interesting now, are you. Everyone's seen that stat of like 95% of AI deployments actually fail where they don't deliver ROI. Now that's been covered a lot. Small sample size, there's a lot of faults with it, but IT is identifying a problem. When it's top down and you're applying AI, IT doesn't work well, as well as bottoms up. If you think about ChatGPT, it wasn't the organizations and the big companies saying, hey, I'm going to make it okay for my employees to ask questions or put sensitive data into the system, right? Like it was employees just saying, is this going to save me 50 minutes is going to save me five minutes. All right, I'm going to dump this in. And they shouldn't have, they should have checked it. But it was kind of that seeing that value that the employees were getting. And then all of a sudden they were like, all right, we got to stop. And if you remember the early days of ChatGPT, they blocked it, they hid the domain, they said no one else could use it. Then people started to use their own laptop to kind of move data over. And so what you saw was this overwhelming force of employees want to adopt it. The enterprises finally said, okay, we've got to do this, we've got to set this up and I think you're going to see that again in the next three to six months. Where we've seen these platforms roll out these AI solutions and they're actually really good. If you only use Google and you want to use the Google Stack, it's great. If you only use Microsoft, only using Microsoft Stack, it's great. But as you start to piece in different areas where you are using Asana, you're using Jira Confluence, your desktop files that are stored on Drive, but you Google Drive, but you actually use Microsoft Outlook for your email, all those things are disparate sources. You're going to see this push to say like I need something and bring all of those things together and enable kind of querying and I can have that central source. And that's what we're really focused in on right now is really being that storage of intelligence and then building automations and workflows from there. And I think that's where the market is really going at the end of the day, I think. You know, like we talked about agents before, like agents are a way to describe something that people don't know what it does yet because people talk about agents all over the place, but really agents are unemployed workers until you find a job for them to actually do something. And if you think about it, once they do something, you're not going to call it agent, you're going to just say it does it just in the same way, like HTML or APIs. No one actually 99% of the people wouldn't know what that is, but it just does the job. And someone's figured out what that layer looks like on top of it.
- [46:28]
B
Yeah, yeah. It kind of struck me because I saw on the Read AI website there's a page there with a bunch of agents that you can enable with your account that do different things. And in many ways, if you're looking at from an AI perspective, it makes sense. You've got all these agents here that you can use. But it was also like, well, these are just things, tasks that people want to accomplish. And in many ways if we hadn't called in an agent, they'd have probably gone into the app and turned on a flipped a switch or something and said, yeah, let me send me a reminder about this thing every day or whatever. So yeah, it's kind of like we're in this weird space right now, which is there's this sort of this heavy AI kind of dominance with everything and you can't ignore that, but there's still a lot of fundamental stuff that doesn't change.
- [47:26]
A
I think you're absolutely right. It doesn't change. And it's really, what problems are you solving with AI versus, like this open world construct? Like, in an open world construct, it's great. You could create an app, you can create any app, but you still need to go in and say, what is that app actually going to do? What problem is it going to solve for me or my team? And what level of efficiency am I going to get out of it? And when we started with the meeting notes side of things, it was really like, nobody wants that job. Like, if you take notes in a meeting and you say, David, you're the one that has to take notes in the meeting. You're like, I am the low person in this meeting because they just want me to take notes on this thing and I got to write it up and I got to send it out. It's a pain if you have to do scheduling. AI should just do the scheduling for you. Especially from a meeting where you've got that automated workflow. It should be able to go in and say, hey, these four people have not come to this meeting in the last two weeks. Let me just pull it from their calendar because it's not important enough from them. And that's stuff that we're doing today where we're pulling out and we're saying, like, hey, just remove these people from a call because they're not showing up, they're showing up late. They're coming in ghost mode where they've got their camera off and they're on mute and they don't talk during that call. And so taking that approach, you're solving problems, you're pushing things in front of their face. And I think if you want to say, like, well, are people really going to give up that level of decision making? Like, is that really going to happen? It's going to be slow initially and we're seeing people start to adopt. But think about it this way. You've got Google Maps and you've got Waze and you've got Apple Maps. The phone is telling you where to go. You've told them what the ultimate goal is. I want to end up at this place. But the phone is actually giving you instruction on what to do. So you're actually kind of a meat bot. You're driving the car, what the phone tells you to do. So I think people will give up that level of control because they're like, yeah, you're going to get me to that spot faster, more efficiently. And that's a fair trade off to say, I don't need to do this anymore.
- [49:14]
B
All right, let's get onto the lightning round. So I've got seven quick fire questions for you. Absolutely. What's one of the best pieces of business advice you've received?
- [49:24]
A
So I was a kid and I was at a client dinner. First client dinner that I had. And I'm Korean, so I'm used to using chopsticks versus like a fork. So I was never taught how to actually correctly hold a fork. So I kind of held it like this, like a conveyor belt kind of thing. And my boss at the time, awesome, he's a mentor, he said, hey, you should actually do it this way. It's very noticeable the way that you do it. And at first I was really offended and I was kind of like, why is he making me look bad? But he took me aside and he told me that. And for me, that taught me a really important business lesson is kind of go be direct. Because there was. He could have tried to do it in a soft way after six or seven or eight dinners, like, hey, David, you know, you're doing it kind of weird way and try to figure out subtly. But the fact that he was direct with me solved a problem for me. And I think that was huge in my career. To go in and just say, like, be direct. If someone has an issue, if something is wrong, don't walk around and try to be kind. Just don't be an either. But be direct.
- [50:19]
B
What book would you recommend to our audience and why?
- [50:22]
A
Made to stick. Great for storytelling. It's a little bit older now. You can probably buy a cheap. But I would highly recommend a quick read. And it really talks about the importance of storytelling. And storytelling is even more important now in the age of 15 second clips. And to take that knowledge and figure out how do I tell a story and a narrative?
- [50:39]
B
What's one attribute or characteristic in your mind of a successful founder?
- [50:43]
A
Always be moving. So for me, if you run into a roadblock, don't lament on it, don't figure out what could have been different, et cetera. Just keep your legs moving. Almost like if you're running back, you're playing football, you want to keep moving because at some point you'll find an opening and you will run through it, but the minute you stop, it's over.
- [51:02]
B
What's your favorite personal productivity tool or habit?
- [51:05]
A
It is delayed send on emails. So people don't use this enough. If someone emails you and you reply right back right away. Most of the time it's a good thing. But sometimes if it's a thoughtful response that they send over, the fact that you respond too quickly makes it look bad. If it's Friday and you're at 9 o' clock at night and you're ready to send something out, don't send it. But you want to kind of say, I'm done mentally. So the way that I work is like I want to check it off a list so I'll delay, send it to say, hey, send it at 8:42am on Monday so they get it. But they don't have to get it during the weekend where they might get lost in the mail.
- [51:35]
B
Love it. What's a new or crazy business idea you'd love to pursue if you had the time?
- [51:40]
A
If I had the time, I would really. I don't have an, I don't have a set idea. But addressing like the homeless crisis, the mental health crisis, I think there's a lot of really smart people that have not tackled that yet. And I think that's one huge opportunity, both for social good but as well as if you're economic good as well, to go in and say, how do I solve this problem that is impacting so many cities, not just in the U.S. but across the globe.
- [52:02]
B
What's an interesting or fun fact about you that most people don't know?
- [52:05]
A
I was the youngest stock investor, youngest investment advisor in the nation when I was 17. So this is where it's like if you run into a roadblock, always keep your legs moving. I was 17 years old. I was trading stocks when this was early days of Internet where every everything had an E in front of it. And the broker that I worked with was E broker. And so I was trading stocks, I was trading stocks too fast as a 17 year old that I broke parts of the system and they reached out and said like, you can't do this anymore. You need to find somebody else to work with. And I was like, how do I get access to these professional systems? And the only way that I could get access if I passed my series 72463 exam and got licensed at both the federal, the organizational sec, nasdaq, all that good stuff as well as the state level. And I did all those tests and they never had a 17 year old do it and then I could start trading stocks and for a while I was doing really well. Where it's like, it was fun.
- [52:54]
B
That is awesome. And finally, what's one of your most important passions outside of your work, it's.
- [52:59]
A
Friends and family, I think, making that. I've never regretted taking a couple extra days off and said like, oh, I need to really get this project done. I remember the times that you hang out that you I got to go see my niece and nephew grow up. I get to see my parents while they're still in good health. Like those experiences, I think the grind is great, but you got to go in and carve that time out because you never worry about, oh, I missed that call like three years ago. It's like, I don't remember who I talked with. I don't even care who I talked with back then. But I do care that I went on vacation with my dad.
- [53:28]
B
David, thanks so much for joining me. It's been a pleasure. If people want to check out Reed AI, they can go to Read AI. And if folks want to get in touch with you, what's the best way for them to do that?
- [53:39]
A
Yeah, easiest way is on LinkedIn. Feel free to send me a note add me. I'm also on Twitter avidshim, so feel free to send me a note there as well.
- [53:48]
B
Sweet. Thanks, man. It's been an absolute pleasure. I just feel like we could have gone on for a couple more hours and probably got a ton more value out of this. So I appreciate you being pretty concise and sharing a ton of wisdom, and I think hopefully a lot of founders listening will walk away with at least one actionable thing that they can apply in their business or just feel a little bit more inspired by hearing your story. So thank you for that.
- [54:17]
A
No, I really appreciate it. This has been a great experience. I really love all the questions that you've asked and really hitting on the stuff where a lot of people skip over. So I think this has been great.
- [54:25]
B
Awesome. I wish you and the team all the best. Thank you. Cheers. Thanks again to Sprinter for supporting this episode. If compliance is eating your team's time, there's a better way. Sprinto automates audits and risk management so you can focus on building, not chasing evidence. SOC 2, ISO 27001, HIPAA, all handled from one simple platform. Go to sprinto.com and book your demo today. That's sprinto.com.